RED-6126: performance-test

*re-enabled overlap detection *re-creating helper document for every page instead of reusing and adding/removing pages
RED-6126: Performance Tests
2023-02-09 11:22:39 +01:00 · 2023-02-09 11:09:52 +01:00 · 2023-02-09 10:57:37 +01:00 · 2023-02-08 16:46:51 +02:00 · 2023-02-07 13:35:32 +01:00 · 2023-02-07 12:52:09 +01:00
43 changed files with 1332 additions and 1075 deletions
@@ -86,16 +86,6 @@
    </dependencies>
    <build>
        <plugins>
-            <plugin>
-                <groupId>org.apache.maven.plugins</groupId>
-                <artifactId>maven-compiler-plugin</artifactId>
-                <configuration>
-                    <annotationProcessors>
-                        <annotationProcessor>lombok.launch.AnnotationProcessorHider$AnnotationProcessor</annotationProcessor>
-                        <annotationProcessor>com.dslplatform.json.processor.CompiledJsonAnnotationProcessor</annotationProcessor>
-                    </annotationProcessors>
-                </configuration>
-            </plugin>
            <plugin>
                <!-- generate git.properties for exposure in /info -->
                <groupId>pl.project13.maven</groupId>
@@ -114,6 +104,16 @@
                    </execution>
                </executions>
            </plugin>
+            <plugin>
+                <groupId>org.apache.maven.plugins</groupId>
+                <artifactId>maven-compiler-plugin</artifactId>
+                <configuration>
+                    <annotationProcessors>
+                        <annotationProcessor>lombok.launch.AnnotationProcessorHider$AnnotationProcessor</annotationProcessor>
+                        <annotationProcessor>com.dslplatform.json.processor.CompiledJsonAnnotationProcessor</annotationProcessor>
+                    </annotationProcessors>
+                </configuration>
+            </plugin>
            <plugin>
                <!-- repackages the generated jar into a runnable fat-jar and makes it
                    executable -->
@@ -1,54 +1,68 @@
 package com.iqser.red.service.ocr.v1.server.model;

-import com.pdftron.pdf.Rect;
-import lombok.Data;
-import lombok.SneakyThrows;
+import static com.iqser.red.service.ocr.v1.server.service.InvisibleElementRemovalService.TOLERANCE;

 import java.awt.geom.Area;
 import java.awt.geom.GeneralPath;
 import java.awt.geom.Rectangle2D;
-import java.util.ArrayDeque;
 import java.util.Deque;
+import java.util.LinkedList;
+
+import com.pdftron.pdf.Rect;
+
+import lombok.Data;
+import lombok.SneakyThrows;

@Data
 public class ClippingPathStack {

-    private Deque<Area> stack = new ArrayDeque<>();
+    private Deque<Area> stack = new LinkedList<>();
+

    @SneakyThrows
    public ClippingPathStack(Rect rectangle) {
+
        stack.push(new Area(new Rectangle2D.Double(rectangle.getX1(), rectangle.getY1(), rectangle.getWidth(), rectangle.getHeight()).getBounds2D()));
    }


    @SneakyThrows
    public void intersectClippingPath(GeneralPath path) {
+
        getCurrentClippingPath().intersect(new Area(path));
    }

+
    public boolean almostIntersects(double x, double y, double width, double height) {
+        // To address inconsistencies in the calculation of the bounding box we slightly increase the rectangle
        // Height or width are zero for straight lines, even though they are being rendered. Therefore, height or width must be at minimum >0.
-        double tolerance = 1e-3;
-        double x_with_tolerance = x > 0 ? x - tolerance : x + tolerance;
-        double y_with_tolerance = y > 0 ? y - tolerance : y + tolerance;
-        double width_with_tolerance = width + 2 * tolerance;
-        double height_with_tolerance = height + 2 * tolerance;
+
+        double x_with_tolerance = x > 0 ? x - TOLERANCE : x + TOLERANCE;
+        double y_with_tolerance = y > 0 ? y - TOLERANCE : y + TOLERANCE;
+        double width_with_tolerance = width + (2 * TOLERANCE);
+        double height_with_tolerance = height + (2 * TOLERANCE);
        return getCurrentClippingPath().intersects(x_with_tolerance, y_with_tolerance, width_with_tolerance, height_with_tolerance);
    }

+
    public Area getCurrentClippingPath() {
+
        return stack.peek();
    }

+
    public void enterNewGState() {
+
        Area current = stack.peek();
        Area cloned = new Area();
        cloned.add(current);
        stack.push(cloned);
    }

+
    public void leaveGState() {
+
        stack.pop();
    }

-}
+}
@@ -1,149 +1,170 @@
 package com.iqser.red.service.ocr.v1.server.model;

+import static com.iqser.red.service.ocr.v1.server.service.InvisibleElementRemovalService.TOLERANCE;
+
+import java.awt.geom.Rectangle2D;
+
 import com.pdftron.common.PDFNetException;
 import com.pdftron.pdf.Element;
 import com.pdftron.pdf.Rect;
-import lombok.*;
+
+import lombok.AccessLevel;
+import lombok.EqualsAndHashCode;
+import lombok.Getter;
+import lombok.SneakyThrows;
+import lombok.experimental.FieldDefaults;
 import lombok.experimental.SuperBuilder;

-import java.awt.geom.Rectangle2D;
-
-@Data
+@Getter
@SuperBuilder
-@NoArgsConstructor
-@AllArgsConstructor
-public abstract class ElementFeatures {
-    private int elementType;
-    private Rectangle2D boundingBox;
+@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
+public class ElementFeatures {
+
+    int elementType;
+    Rectangle2D boundingBox;
+

    public boolean almostMatches(Element element) throws PDFNetException {
-        if (element.getType() != elementType) return false;
-        if (element.getBBox() == null) return false;
-        return rectsAlmostMatch(element.getBBox());
+
+        return element.getType() == elementType && //
+                element.getBBox() != null && //
+                rectsAlmostMatch(element.getBBox());
    }

+
    protected boolean almostEqual(double a, double b) {
-        double tolerance = 1e-3;
-        return Math.abs(a - b) < tolerance;
+
+        return Math.abs(a - b) < TOLERANCE;
    }

+
    @SneakyThrows
    private boolean rectsAlmostMatch(Rect bBox) {
-        if (!almostEqual(bBox.getX1(), boundingBox.getX())) return false;
-        if (!almostEqual(bBox.getY1(), boundingBox.getY())) return false;
-        if (!almostEqual(bBox.getWidth(), boundingBox.getWidth())) return false;
-        return almostEqual(bBox.getHeight(), boundingBox.getHeight());
+        // To address the inconsistencies in the calculation of the bounding box we check equality with a tolerance
+
+        return almostEqual(bBox.getX1(), boundingBox.getX()) && //
+                almostEqual(bBox.getY1(), boundingBox.getY()) && //
+                almostEqual(bBox.getWidth(), boundingBox.getWidth()) && //
+                almostEqual(bBox.getHeight(), boundingBox.getHeight());
    }

+
    @EqualsAndHashCode(callSuper = true)
-    @Data
+    @Getter
    @SuperBuilder
-    @NoArgsConstructor
-    @AllArgsConstructor
-    public static class Text extends ElementFeatures {
-        private String text;
-        private int font;
-        private double fontsize;
+    @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
+    private static class Text extends ElementFeatures {
+
+        String text;
+        int font;
+        double fontsize;
+

        @Override
        public boolean almostMatches(Element element) throws PDFNetException {
-            if (!super.almostMatches(element)) return false;
-            if (!text.equals(element.getTextString())) return false;
-            if (font != element.getGState().getFont().getType()) return false;
-            return almostEqual(fontsize, element.getGState().getFontSize());
+
+            return super.almostMatches(element) && //
+                    text.equals(element.getTextString()) && //
+                    font == element.getGState().getFont().getType() && //
+                    almostEqual(fontsize, element.getGState().getFontSize());
        }

    }

    @EqualsAndHashCode(callSuper = true)
-    @Data
+    @Getter
    @SuperBuilder
-    @NoArgsConstructor
-    @AllArgsConstructor
-    public static class Path extends ElementFeatures {
-        private boolean isClippingPath;
-        private boolean isClipWindingFill;
-        private boolean isStroked;
-        private boolean isFilled;
-        private boolean isWindingFill;
+    @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
+    private static class Path extends ElementFeatures {
+
+        boolean isClippingPath;
+        boolean isClipWindingFill;
+        boolean isStroked;
+        boolean isFilled;
+        boolean isWindingFill;
+

        @Override
        public boolean almostMatches(Element element) throws PDFNetException {
-            if (!super.almostMatches(element)) return false;
-            if (isClippingPath != element.isClippingPath()) return false;
-            if (isClipWindingFill != element.isClipWindingFill()) return false;
-            if (isStroked != element.isStroked()) return false;
-            if (isFilled != element.isFilled()) return false;
-            if (isWindingFill != element.isWindingFill()) return false;

-            return true;
+            return super.almostMatches(element) && //
+                    isClippingPath == element.isClippingPath() && //
+                    isClipWindingFill == element.isClipWindingFill() && //
+                    isStroked == element.isStroked() && //
+                    isFilled == element.isFilled() && //
+                    isWindingFill == element.isWindingFill();
+
        }
+
    }

    @EqualsAndHashCode(callSuper = true)
-    @Data
+    @Getter
    @SuperBuilder
-    @NoArgsConstructor
-    public static class Image extends ElementFeatures {
-        private int dataSize;
-        private int height;
-        private int width;
-        private int renderingIntent;
-        private int componentNum;
-        private int bitsPerComponent;
+    @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
+    private static class Image extends ElementFeatures {
+
+        int dataSize;
+        int height;
+        int width;
+        int renderingIntent;
+        int componentNum;
+        int bitsPerComponent;
+

        @Override
        public boolean almostMatches(Element element) throws PDFNetException {
-            if (!super.almostMatches(element)) return false;
-            if (dataSize != element.getImageDataSize()) return false;
-            if (height != element.getImageHeight()) return false;
-            if (width != element.getImageWidth()) return false;
-            if (renderingIntent != element.getImageRenderingIntent()) return false;
-            if (componentNum != element.getComponentNum()) return false;
-            if (bitsPerComponent != element.getBitsPerComponent()) return false;
-            return true;
+
+            return super.almostMatches(element) && //
+                    dataSize == element.getImageDataSize() && //
+                    height == element.getImageHeight() && //
+                    width == element.getImageWidth() && //
+                    renderingIntent == element.getImageRenderingIntent() && //
+                    componentNum == element.getComponentNum() && //
+                    bitsPerComponent == element.getBitsPerComponent();
        }
+
    }

+
    public static ElementFeatures extractFeatures(Element element) throws PDFNetException {
-        switch (element.getType()) {
-            case Element.e_path:
-                return ElementFeatures.Path.builder()
-                        .elementType(element.getType())
-                        .boundingBox(toRectangle2D(element.getBBox()))
-                        .isClippingPath(element.isClippingPath())
-                        .isClipWindingFill(element.isClipWindingFill())
-                        .isStroked(element.isStroked())
-                        .isFilled(element.isFilled())
-                        .isWindingFill(element.isWindingFill())
-                        .build();
-            case Element.e_text:
-                return ElementFeatures.Text.builder()
-                        .elementType(element.getType())
-                        .boundingBox(toRectangle2D(element.getBBox()))
-                        .text(element.getTextString())
-                        .font(element.getGState().getFont().getType())
-                        .fontsize(element.getGState().getFontSize())
-                        .build();
-            case Element.e_image:
-            case Element.e_inline_image:
-                return Image.builder()
-                        .elementType(element.getType())
-                        .boundingBox(toRectangle2D(element.getBBox()))
-                        .dataSize(element.getImageDataSize())
-                        .height(element.getImageHeight())
-                        .width(element.getImageWidth())
-                        .renderingIntent(element.getImageRenderingIntent())
-                        .componentNum(element.getComponentNum())
-                        .bitsPerComponent(element.getBitsPerComponent())
-                        .build();
-            default:
-                throw new UnsupportedOperationException("Feature Extraction is not supported for PDFTron.Element with type: " + element.getType());
-        }
+
+        return switch (element.getType()) {
+            case Element.e_path -> Path.builder()
+                    .elementType(element.getType())
+                    .boundingBox(toRectangle2D(element.getBBox()))
+                    .isClippingPath(element.isClippingPath())
+                    .isClipWindingFill(element.isClipWindingFill())
+                    .isStroked(element.isStroked())
+                    .isFilled(element.isFilled())
+                    .isWindingFill(element.isWindingFill())
+                    .build();
+            case Element.e_text -> Text.builder()
+                    .elementType(element.getType())
+                    .boundingBox(toRectangle2D(element.getBBox()))
+                    .text(element.getTextString())
+                    .font(element.getGState().getFont().getType())
+                    .fontsize(element.getGState().getFontSize())
+                    .build();
+            case Element.e_image, Element.e_inline_image -> Image.builder()
+                    .elementType(element.getType())
+                    .boundingBox(toRectangle2D(element.getBBox()))
+                    .dataSize(element.getImageDataSize())
+                    .height(element.getImageHeight())
+                    .width(element.getImageWidth())
+                    .renderingIntent(element.getImageRenderingIntent())
+                    .componentNum(element.getComponentNum())
+                    .bitsPerComponent(element.getBitsPerComponent())
+                    .build();
+            // This technically should never happen, it's a safetynet
+            default -> throw new RuntimeException("Feature Extraction is not supported for PDFTron.Element with type: " + element.getType());
+        };
    }

+
    private static Rectangle2D toRectangle2D(Rect rect) throws PDFNetException {
+
        return new Rectangle2D.Double(rect.getX1(), rect.getY1(), rect.getWidth(), rect.getHeight());
    }
-}

+}
@@ -0,0 +1,177 @@
+package com.iqser.red.service.ocr.v1.server.service;
+
+import java.awt.geom.Point2D;
+import java.awt.geom.Rectangle2D;
+import java.awt.geom.RectangularShape;
+import java.util.Comparator;
+import java.util.HashMap;
+import java.util.LinkedList;
+import java.util.List;
+import java.util.Map;
+
+import org.springframework.stereotype.Service;
+
+import com.pdftron.common.Matrix2D;
+import com.pdftron.common.PDFNetException;
+import com.pdftron.pdf.Element;
+import com.pdftron.pdf.ElementReader;
+import com.pdftron.pdf.PDFDoc;
+import com.pdftron.pdf.Page;
+import com.pdftron.pdf.Rect;
+import com.pdftron.pdf.RectCollection;
+
+import lombok.SneakyThrows;
+
+@Service
+public class ImagePositionRetrievalService {
+
+    private static final double TOLERANCE = 1e-1;
+
+
+    /**
+     * Iterates over all elements in a PDF Document and retrieves the bounding box for each image,
+     * Then it adjusts the bounding boxes for the page rotation.
+     * If the mirrorY flag is set, the Y Coordinates are mirrored and moved up by the page height. This is required for PDFTrons OCRModule.
+     *
+     * @param pdfDoc  a PDF File as PDFTron PDFDoc class
+     * @param mirrorY if this flag is set, all coordinates are calculated with upper left corner as (0,0), else initial user space
+     * @return a map with the page indices as keys and the image bounding boxes on that page as a RectCollection
+     */
+    @SneakyThrows
+    public Map<Integer, RectCollection> getImagePositionPerPage(PDFDoc pdfDoc, boolean mirrorY) {
+
+        Map<Integer, RectCollection> pageIdToImagePositions = new HashMap<>();
+        ElementReader reader = new ElementReader();
+        for (int pageId = 1; pageId <= pdfDoc.getPageCount(); ++pageId) {
+            RectCollection imagePositions = new RectCollection();
+            reader.begin(pdfDoc.getPage(pageId));
+
+            findImagePositionsOnPage(reader, imagePositions, pdfDoc.getPage(pageId), mirrorY);
+            imagePositions = mergeOverlappingRects(imagePositions);
+
+            reader.end();
+            if (imagePositions.getNumRects() > 0) {
+                pageIdToImagePositions.put(pageId, imagePositions);
+            }
+        }
+        return pageIdToImagePositions;
+    }
+
+
+    private void findImagePositionsOnPage(ElementReader reader, RectCollection imagePositions, Page currentPage, boolean mirrorY) throws PDFNetException {
+
+        Element element;
+        while ((element = reader.next()) != null) {
+            switch (element.getType()) {
+                case Element.e_image, Element.e_inline_image -> imagePositions.addRect(toRotationAdjustedRect(element.getBBox(), currentPage, mirrorY));
+                case Element.e_form -> {
+                    reader.formBegin();
+                    findImagePositionsOnPage(reader, imagePositions, currentPage, mirrorY);
+                    reader.end();
+                }
+            }
+        }
+    }
+
+
+    @SneakyThrows
+    public RectCollection mergeOverlappingRects(RectCollection imagePositions) {
+
+        if (imagePositions.getNumRects() == 1) {
+            return imagePositions;
+        }
+
+        List<Rectangle2D> rectangleList = toSortedRectangleList(imagePositions);
+
+        rectangleList = mergeRectangleListRecursive(rectangleList, 0);
+        return toRectCollection(rectangleList);
+    }
+
+
+    // Sometimes images are split up into stripes, here we merge the positions of aligned and intersecting rectangles into one larger rectangle
+    private List<Rectangle2D> mergeRectangleListRecursive(List<Rectangle2D> rectangleList, int currentIdx) {
+
+        if (rectangleList.size() < currentIdx + 2) {
+            return rectangleList;
+        }
+
+        var rect1 = rectangleList.get(currentIdx);
+        var rect2 = rectangleList.get(currentIdx + 1);
+
+        boolean isAlignedX = Math.abs(rect1.getMinX() - rect2.getMinX()) < TOLERANCE && Math.abs(rect1.getMaxX() - rect2.getMaxX()) < TOLERANCE;
+        boolean isAlignedY = Math.abs(rect1.getMinY() - rect2.getMinY()) < TOLERANCE && Math.abs(rect1.getMaxY() - rect2.getMaxY()) < TOLERANCE;
+        boolean intersects = rect1.intersects(rect2.getMinX() - TOLERANCE, rect2.getMinY() - TOLERANCE, rect2.getWidth() + (2 * TOLERANCE), rect2.getHeight() + (2 * TOLERANCE));
+
+        if (intersects && (isAlignedX || isAlignedY)) {
+            rectangleList.remove(currentIdx + 1);
+            rectangleList.remove(currentIdx);
+            rectangleList.add(currentIdx, rect1.createUnion(rect2));
+            return mergeRectangleListRecursive(rectangleList, currentIdx);
+        } else {
+            return mergeRectangleListRecursive(rectangleList, currentIdx + 1);
+        }
+    }
+
+
+    private Rect toRotationAdjustedRect(Rect bbox, Page page, boolean mirrorY) throws PDFNetException {
+
+        int rotation = page.getRotation();
+        double height = page.getPageHeight();
+        double width = page.getPageWidth();
+
+        // Even though PDFTron almost always has the origin in the lower left corner, for some reason, the OCRModule's addTextZonesForPage() uses the upper left corner as origin...
+        Matrix2D mirrorMatrix;
+        if (mirrorY) {
+            mirrorMatrix = new Matrix2D(1, 0, 0, -1, 0, height);
+        } else {
+            mirrorMatrix = new Matrix2D();
+        }
+
+        // We need to rotate the rects to fit to the page rotation
+        Matrix2D rotationMatrix = switch (rotation) {
+            case 1 -> new Matrix2D(0, -1, 1, 0, 0, height);
+            case 2 -> new Matrix2D(-1, 0, 0, -1, width, height);
+            case 3 -> new Matrix2D(0, 1, -1, 0, width, 0);
+            default -> new Matrix2D();
+        };
+
+        Matrix2D finalMatrix = mirrorMatrix.multiply(rotationMatrix);
+
+        Point2D.Double p1 = finalMatrix.multPoint(bbox.getX1(), bbox.getY1());
+        Point2D.Double p2 = finalMatrix.multPoint(bbox.getX2(), bbox.getY2());
+
+        // PDFTron Rect *needs* lower left and upper right coordinates to calculate width and height correctly, even though the documentation states otherwise
+        Point2D.Double lowerLeft = new Point2D.Double(Math.min(p1.x, p2.x), Math.min(p1.y, p2.y));
+        Point2D.Double upperRight = new Point2D.Double(Math.max(p1.x, p2.x), Math.max(p1.y, p2.y));
+
+        return new Rect(lowerLeft.x, lowerLeft.y, upperRight.x, upperRight.y);
+    }
+
+
+    private RectCollection toRectCollection(List<Rectangle2D> rectangleList) {
+
+        RectCollection rectCollection = new RectCollection();
+        rectangleList.forEach(r -> {
+            try {
+                rectCollection.addRect(new Rect(r.getMinX(), r.getMinY(), r.getMaxX(), r.getMaxY()));
+            } catch (PDFNetException e) {
+                throw new RuntimeException(e);
+            }
+        });
+        return rectCollection;
+    }
+
+
+    @SneakyThrows
+    private List<Rectangle2D> toSortedRectangleList(RectCollection rectCollection) {
+
+        List<Rectangle2D> list = new LinkedList<>();
+        for (int i = 0; i < rectCollection.getNumRects(); ++i) {
+            Rect r = rectCollection.getRectAt(i);
+            list.add(new Rectangle2D.Double(r.getX1(), r.getY1(), r.getWidth(), r.getHeight()));
+        }
+        list.sort(Comparator.comparingDouble(RectangularShape::getMinY).thenComparing(RectangularShape::getMinX));
+        return list;
+    }
+
+}
@@ -0,0 +1,442 @@
+package com.iqser.red.service.ocr.v1.server.service;
+
+import java.awt.Shape;
+import java.awt.geom.AffineTransform;
+import java.awt.geom.GeneralPath;
+import java.awt.geom.Rectangle2D;
+import java.io.InputStream;
+import java.io.OutputStream;
+import java.util.ArrayList;
+import java.util.Iterator;
+import java.util.List;
+import java.util.Set;
+import java.util.TreeSet;
+
+import org.springframework.stereotype.Service;
+
+import com.google.common.primitives.Bytes;
+import com.google.common.primitives.Doubles;
+import com.iqser.red.service.ocr.v1.server.model.ClippingPathStack;
+import com.iqser.red.service.ocr.v1.server.model.ElementFeatures;
+import com.pdftron.common.Matrix2D;
+import com.pdftron.common.PDFNetException;
+import com.pdftron.pdf.ColorPt;
+import com.pdftron.pdf.ColorSpace;
+import com.pdftron.pdf.Element;
+import com.pdftron.pdf.ElementBuilder;
+import com.pdftron.pdf.ElementReader;
+import com.pdftron.pdf.ElementWriter;
+import com.pdftron.pdf.GState;
+import com.pdftron.pdf.PDFDoc;
+import com.pdftron.pdf.Page;
+import com.pdftron.pdf.PageIterator;
+import com.pdftron.pdf.PathData;
+import com.pdftron.pdf.Rect;
+import com.pdftron.sdf.Obj;
+import com.pdftron.sdf.SDFDoc;
+
+import lombok.Builder;
+import lombok.SneakyThrows;
+import lombok.extern.slf4j.Slf4j;
+
+@Slf4j
+@Service
+public class InvisibleElementRemovalService {
+
+    static public final double TOLERANCE = 1e-3;
+
+
+    /**
+     * Removes all hidden Text, Path and Image Elements from a PDF Document.
+     * handled cases:
+     * -Text which is transparent or is set to not render
+     * -Elements outside of clipping path
+     * -Elements that have been painted over by visible and filled Paths
+     * unhandled cases:
+     * -Elements covered by widely stroked path
+     * -Elements with the same color as background
+     * -Any Text set to clipping with its many interactions with other elements
+     *
+     * @param pdfFile The PDF file to process
+     * @param delta   If this flag is set only the removed Elements will be written to the output file.
+     *                The Elements are red if they are removed by clipping path, blue for transparency, and a green bounding box for overlap.
+     * @param out     OutputStream to write the resulting file to
+     **/
+    @SneakyThrows
+    public void removeInvisibleElements(InputStream pdfFile, OutputStream out, boolean delta) {
+
+        PDFDoc pdfDoc = new PDFDoc(pdfFile);
+
+        ElementWriter writer = new ElementWriter();
+        ElementReader reader = new ElementReader();
+        Set<Long> visitedXObjIds = new TreeSet<>();
+
+        for (PageIterator iterator = pdfDoc.getPageIterator(); iterator.hasNext(); ) {
+
+            Page page = iterator.next();
+
+            visitedXObjIds.add(page.getSDFObj().getObjNum());
+            InvisibleElementRemovalContext context = InvisibleElementRemovalContext.builder()
+                    .reader(reader)
+                    .clippingPathStack(new ClippingPathStack(page.getMediaBox()))
+                    .delta(delta)
+                    .overlappedElements(new ArrayList<>())
+                    .visibleElements(new ArrayList<>())
+                    .visitedXObjIds(visitedXObjIds)
+                    .build();
+
+            removeClippedElementsAndInvisibleTextAndRememberOverlappedElements(page, writer, context);
+
+            context.visitedXObjIds().clear();
+
+            removeOverlappedElements(page, writer, context);
+        }
+        pdfDoc.save(out, SDFDoc.SaveMode.REMOVE_UNUSED, null);
+    }
+
+
+    private void removeClippedElementsAndInvisibleTextAndRememberOverlappedElements(Page page,
+                                                                                    ElementWriter writer,
+                                                                                    InvisibleElementRemovalContext context) throws PDFNetException {
+
+        context.reader().begin(page);
+        writer.begin(page, ElementWriter.e_replacement, false, true, page.getResourceDict());
+        processElements(writer, context);
+        writer.end();
+        context.reader().end();
+    }
+
+
+    private void processElements(ElementWriter writer, InvisibleElementRemovalContext context) throws PDFNetException {
+
+        for (Element element = context.reader().next(); element != null; element = context.reader().next())
+            switch (element.getType()) {
+                case Element.e_image, Element.e_inline_image -> processImages(element, writer, context);
+                case Element.e_text -> processText(element, writer, context);
+                case Element.e_path -> processPath(element, writer, context);
+                case Element.e_form -> processForm(element, writer, context);
+                case Element.e_group_begin -> {
+                    context.clippingPathStack().enterNewGState();
+                    writer.writeElement(element);
+                }
+                case Element.e_group_end -> {
+                    context.clippingPathStack().leaveGState();
+                    writer.writeElement(element);
+                }
+                default -> writer.writeElement(element);
+            }
+    }
+
+
+    private void processImages(Element imageElement, ElementWriter writer, InvisibleElementRemovalContext context) throws PDFNetException {
+
+        Rect rect = imageElement.getBBox();
+
+        if (rect == null) {
+            return;
+        }
+
+        boolean inClippingPath = context.clippingPathStack().almostIntersects(rect.getX1(), rect.getY1(), rect.getWidth(), rect.getHeight());
+
+        if (!context.delta() && inClippingPath) {
+            context.visibleElements().add(ElementFeatures.extractFeatures(imageElement));
+        }
+
+        if (context.delta() ^ inClippingPath) {
+            writer.writeElement(imageElement);
+        }
+    }
+
+
+    private void processText(Element textElement, ElementWriter writer, InvisibleElementRemovalContext context) throws PDFNetException {
+
+        Rect rect = textElement.getBBox();
+
+        if (rect == null) {
+            writer.writeElement(textElement);
+            return;
+        }
+
+        GState gState = textElement.getGState();
+
+        boolean inClippingPath = context.clippingPathStack().almostIntersects(rect.getX1(), rect.getY1(), rect.getWidth(), rect.getHeight());
+
+        boolean isTextVisible = isTextRenderedVisibly(gState);
+
+        if (inClippingPath && isTextVisible) {
+            context.visibleElements().add(ElementFeatures.extractFeatures(textElement));
+        }
+        if (!context.delta()) {
+            if (inClippingPath && isTextVisible) {
+                writer.writeElement(textElement);
+            } else if (textElement.hasTextMatrix()) {
+                /*
+                PDFTron Element with type "text" refers to a Tj command. If a Tm command is just above it in the pdf file, PDFTron will join the two commands and treat them as one Element.
+                hasTextMatrix() checks for this case specifically. Also, Tm changes the position for a whole BT/ET segment, possibly containing multiple Tj commands.
+                Therefore, the position of a following Tj is affected by not writing the first Element.
+                This is why, we write only the Tm command:
+                */
+                writer.writeGStateChanges(textElement);
+            }
+        } else {
+            if (!inClippingPath) {
+                gState.setFillColorSpace(ColorSpace.createDeviceRGB());
+                // red for elements removed by clipping path
+                gState.setFillColor(new ColorPt(1, 0, 0));
+                writer.writeElement(textElement);
+            }
+            if (!isTextVisible) {
+                gState.setFillColorSpace(ColorSpace.createDeviceRGB());
+                // blue for elements removed due to transparency or not rendered
+                gState.setFillColor(new ColorPt(0, 0, 1));
+                gState.setTextRenderMode(GState.e_fill_text);
+                gState.setFillOpacity(1);
+                writer.writeElement(textElement);
+            }
+        }
+    }
+
+
+    private void processForm(Element formElement, ElementWriter writer, InvisibleElementRemovalContext context) throws PDFNetException {
+
+        writer.writeElement(formElement);
+        Obj formObj = formElement.getXObject();
+
+        if (!context.visitedXObjIds().contains(formObj.getObjNum())) {
+            context.visitedXObjIds().add(formObj.getObjNum());
+            // writer needs to be newly initialized when entering a new content stream
+            // see ElementEditTest in PDFTron (https://www.pdftron.com/documentation/samples/android/java/ElementEditTest)
+            ElementWriter formWriter = new ElementWriter();
+            context.reader().formBegin();
+            formWriter.begin(formObj);
+
+            context.reader().clearChangeList();
+            formWriter.setDefaultGState(context.reader());
+
+            processElements(formWriter, context);
+            formWriter.end();
+            context.reader().end();
+        }
+    }
+
+
+    private void processPath(Element pathElement, ElementWriter writer, InvisibleElementRemovalContext context) throws PDFNetException {
+
+        GeneralPath linePath = convertToGeneralPath(pathElement.getPathData());
+
+        //transform path to initial user space
+        var ctm = pathElement.getCTM();
+        var affineTransform = toAffineTransform(ctm);
+        linePath.transform(affineTransform);
+
+        var rect = linePath.getBounds2D();
+
+        boolean inClippingPath = context.clippingPathStack().almostIntersects(rect.getX(), rect.getY(), rect.getWidth(), rect.getHeight());
+
+        if (pathElement.isClippingPath()) {
+            if (pathElement.isClipWindingFill()) {
+                linePath.setWindingRule(GeneralPath.WIND_NON_ZERO);
+            } else {
+                linePath.setWindingRule(GeneralPath.WIND_EVEN_ODD);
+            }
+
+            context.clippingPathStack().intersectClippingPath(linePath);
+            pathElement.setPathClip(!context.delta());
+            writer.writeElement(pathElement);
+
+        } else {
+            if (pathElement.isWindingFill()) {
+                linePath.setWindingRule(GeneralPath.WIND_NON_ZERO);
+            } else {
+                linePath.setWindingRule(GeneralPath.WIND_EVEN_ODD);
+            }
+
+            if (inClippingPath) {
+                if (isFilledAndNonTransparent(pathElement)) {
+                    List<ElementFeatures> currentOverlappedElements = context.visibleElements()
+                            .stream()
+                            .filter(features -> almostContains(linePath, features.getBoundingBox()))
+                            .toList();
+                    context.overlappedElements().addAll(currentOverlappedElements);
+                    context.visibleElements().removeAll(currentOverlappedElements);
+                }
+                context.visibleElements().add(ElementFeatures.extractFeatures(pathElement));
+                if (!context.delta()) {
+                    writer.writeElement(pathElement);
+                }
+            }
+            if (context.delta() && !inClippingPath) {
+                pathElement.getGState().setFillColorSpace(ColorSpace.createDeviceRGB());
+                pathElement.getGState().setFillColor(new ColorPt(1, 0, 0));
+                pathElement.getGState().setStrokeColorSpace(ColorSpace.createDeviceRGB());
+                pathElement.getGState().setStrokeColor(new ColorPt(1, 0, 0));
+                writer.writeElement(pathElement);
+            }
+        }
+    }
+
+
+    private void removeOverlappedElements(Page page, ElementWriter writer, InvisibleElementRemovalContext context) throws PDFNetException {
+
+        context.reader().begin(page);
+        writer.begin(page, ElementWriter.e_replacement, false, true, page.getResourceDict());
+        if (context.delta()) {
+            // green for element removed due to overlapping
+            context.overlappedElements().forEach(feature -> drawBBox(writer, feature.getBoundingBox(), "#00FF00"));
+            context.overlappedElements().clear();
+        }
+        processOverlappedElements(writer, context);
+        writer.end();
+        context.reader().end();
+
+        if (context.overlappedElements().size() > 0) {
+            log.warn(context.overlappedElements().size() + " overlapped elements have not been found or removed");
+        }
+    }
+
+
+    private void processOverlappedElements(ElementWriter writer, InvisibleElementRemovalContext context) throws PDFNetException {
+
+        for (Element element = context.reader().next(); element != null; element = context.reader().next()) {
+            switch (element.getType()) {
+                case Element.e_form -> processFormOverlappedElements(writer, element, context);
+                case Element.e_path, Element.e_image, Element.e_inline_image, Element.e_text -> {
+                    boolean anyMatch = false;
+                    for (ElementFeatures elementToRemove : context.overlappedElements()) {
+                        if (elementToRemove.almostMatches(element)) {
+                            context.overlappedElements().remove(elementToRemove);
+                            anyMatch = true;
+                            break;
+                        }
+                    }
+                    if (!anyMatch) {
+                        writer.writeElement(element);
+                    } else if (element.getType() == 3 && element.hasTextMatrix()) {
+                        /*
+                        PDFTron Element with type "text" refers to a Tj command. If a Tm command is just above it in the pdf file, PDFTron will join the two commands and treat them as one Element.
+                        hasTextMatrix() checks for this case specifically. Also, Tm changes the position for a whole BT/ET segment, possibly containing multiple Tj commands.
+                        Therefore, the position of a following Tj is affected by not writing the first Element.
+                        This is why, we write only the Tm command:
+                        */
+                        writer.writeGStateChanges(element);
+                    }
+                }
+                default -> writer.writeElement(element);
+            }
+        }
+    }
+
+
+    private void processFormOverlappedElements(ElementWriter writer, Element formElement, InvisibleElementRemovalContext context) throws PDFNetException {
+
+        writer.writeElement(formElement);
+        Obj formObj = formElement.getXObject();
+
+        if (!context.visitedXObjIds().contains(formObj.getObjNum())) {
+            context.visitedXObjIds().add(formObj.getObjNum());
+            // writer needs to be newly initialized when entering a new content stream
+            // see ElementEditTest in PDFTron (https://www.pdftron.com/documentation/samples/android/java/ElementEditTest)
+            ElementWriter formWriter = new ElementWriter();
+            context.reader().formBegin();
+            formWriter.begin(formObj);
+
+            context.reader().clearChangeList();
+            formWriter.setDefaultGState(context.reader());
+
+            processOverlappedElements(formWriter, context);
+            formWriter.end();
+            context.reader().end();
+        }
+    }
+
+
+    private boolean isTextRenderedVisibly(GState gState) throws PDFNetException {
+
+        return gState.getTextRenderMode() != GState.e_invisible_text && //
+                !(gState.getTextRenderMode() == GState.e_fill_text && gState.getFillOpacity() == 0) && //
+                !(gState.getTextRenderMode() == GState.e_stroke_text && gState.getStrokeOpacity() == 0) && //
+                !(gState.getTextRenderMode() == GState.e_fill_stroke_text && gState.getFillOpacity() == 0 && gState.getStrokeOpacity() == 0);
+    }
+
+
+    private GeneralPath convertToGeneralPath(PathData pathData) throws PDFNetException {
+
+        GeneralPath linePath = new GeneralPath();
+        Iterator<Double> points = Doubles.asList(pathData.getPoints()).iterator();
+        Iterable<Byte> operators = Bytes.asList(pathData.getOperators());
+        for (var operator : operators) {
+            switch (operator) {
+                case PathData.e_moveto -> linePath.moveTo(points.next(), points.next());
+                case PathData.e_lineto -> linePath.lineTo(points.next(), points.next());
+                case PathData.e_cubicto -> linePath.curveTo(points.next(), points.next(), points.next(), points.next(), points.next(), points.next());
+                case PathData.e_closepath -> linePath.closePath();
+                case PathData.e_rect -> {
+                    double x = points.next();
+                    double y = points.next();
+                    double w = points.next();
+                    double h = points.next();
+                    linePath.moveTo(x, y);
+                    linePath.lineTo(x + w, y);
+                    linePath.lineTo(x + w, y + h);
+                    linePath.lineTo(x, y + h);
+                    linePath.closePath();
+                }
+                default -> throw new PDFNetException("Invalid Element Type", 0, "", "", "");
+            }
+        }
+        return linePath;
+    }
+
+
+    private boolean almostContains(Shape outer, Rectangle2D inner) {
+        //To address inconsistencies in the calculation of the bounding box we slightly shrink the inner rectangle
+
+        double x_with_tolerance = inner.getX() >= 0 ? inner.getX() + TOLERANCE : inner.getX() - TOLERANCE;
+        double y_with_tolerance = inner.getY() >= 0 ? inner.getY() + TOLERANCE : inner.getY() - TOLERANCE;
+        double height_with_tolerance = inner.getHeight() - (2 * TOLERANCE);
+        double width_with_tolerance = inner.getWidth() - (2 * TOLERANCE);
+        Rectangle2D innerRect = new Rectangle2D.Double(x_with_tolerance, y_with_tolerance, width_with_tolerance, height_with_tolerance);
+
+        return outer.contains(innerRect);
+    }
+
+
+    private boolean isFilledAndNonTransparent(Element element) throws PDFNetException {
+
+        return element.isFilled() && element.getGState().getFillOpacity() == 1;
+    }
+
+
+    @SneakyThrows
+    private void drawBBox(ElementWriter writer, Rectangle2D r, String hexcolor) {
+
+        ColorPt colorPt = new ColorPt(Integer.valueOf(hexcolor.substring(1, 3), 16) / 255d,
+                Integer.valueOf(hexcolor.substring(3, 5), 16) / 255d,
+                Integer.valueOf(hexcolor.substring(5, 7), 16) / 255d);
+        ElementBuilder eb = new ElementBuilder();
+        Element rect = eb.createRect(r.getX(), r.getY(), r.getWidth(), r.getHeight());
+        rect.setPathStroke(true);
+        rect.getGState().setStrokeColorSpace(ColorSpace.createDeviceRGB());
+        rect.getGState().setStrokeColor(colorPt);
+        writer.writePlacedElement(rect);
+    }
+
+
+    private static AffineTransform toAffineTransform(Matrix2D ctm) throws PDFNetException {
+
+        return new AffineTransform(ctm.getA(), ctm.getB(), ctm.getC(), ctm.getD(), ctm.getH(), ctm.getV());
+    }
+
+
+    @Builder
+    private record InvisibleElementRemovalContext(
+            boolean delta,
+            ElementReader reader,
+            ClippingPathStack clippingPathStack,
+            List<ElementFeatures> overlappedElements,
+            List<ElementFeatures> visibleElements,
+            Set<Long> visitedXObjIds) {
+
+    }
+
+}
@@ -1,463 +0,0 @@
-package com.iqser.red.service.ocr.v1.server.service;
-
-import com.iqser.red.service.ocr.v1.server.model.ClippingPathStack;
-import com.iqser.red.service.ocr.v1.server.model.ElementFeatures;
-import com.pdftron.common.Matrix2D;
-import com.pdftron.common.PDFNetException;
-import com.pdftron.pdf.*;
-import com.pdftron.sdf.Obj;
-import com.pdftron.sdf.SDFDoc;
-import lombok.SneakyThrows;
-import lombok.extern.slf4j.Slf4j;
-import org.apache.commons.io.FileUtils;
-import org.springframework.stereotype.Service;
-
-import java.awt.*;
-import java.awt.geom.AffineTransform;
-import java.awt.geom.GeneralPath;
-import java.awt.geom.Rectangle2D;
-import java.io.File;
-import java.io.IOException;
-import java.util.ArrayList;
-import java.util.List;
-import java.util.Set;
-import java.util.TreeSet;
-import java.util.stream.Collectors;
-
-@Slf4j
-@Service
-public class InvisibleElementService {
-
-    /*
-    handled cases:
-        Text or Path outside of clipping path
-        Text which is transparent or is set to not render
-        Text or Path that have been painted over by visible and filled Paths
-    unhandled cases:
-        Text covered by widely stroked path
-        Text same color as background
-        Any Text set to clipping with its many interactions with other elements
-     */
-    @SneakyThrows
-    public byte[] removeInvisibleTextOrPathElements(byte[] pdfFile, boolean delta) {
-        PDFDoc pdfDoc = new PDFDoc(pdfFile);
-
-        ElementWriter writer = new ElementWriter();
-        ElementReader reader = new ElementReader();
-        Set<Integer> visited = new TreeSet<>();
-
-        for (PageIterator iterator = pdfDoc.getPageIterator(); iterator.hasNext(); ) {
-            Page page = iterator.next();
-            List<ElementFeatures> overlappedElements = removeClippedElementsAndInvisibleTextAndRememberOverlappedElements(page, reader, writer, visited, delta);
-            visited.clear();
-            removeOverlappedElements(page, reader, writer, visited, overlappedElements, delta);
-        }
-
-        if (delta) {
-            debugSave(pdfDoc.save(SDFDoc.SaveMode.LINEARIZED, null));
-        }
-
-        return pdfDoc.save(SDFDoc.SaveMode.LINEARIZED, null);
-    }
-
-    private List<ElementFeatures> removeClippedElementsAndInvisibleTextAndRememberOverlappedElements(Page page, ElementReader reader, ElementWriter writer, Set<Integer> visited, boolean delta)
-            throws PDFNetException {
-        var overlappedElements = new ArrayList<ElementFeatures>();
-        var visibleElements = new ArrayList<ElementFeatures>();
-        ClippingPathStack clippingPathStack = new ClippingPathStack(page.getMediaBox());
-        visited.add((int) page.getSDFObj().getObjNum());
-        reader.begin(page);
-        writer.begin(page, ElementWriter.e_replacement, false, true, page.getResourceDict());
-        processElements(reader, writer, visited, clippingPathStack, delta, overlappedElements, visibleElements);
-        writer.end();
-        reader.end();
-        return overlappedElements;
-    }
-
-    private void processElements(ElementReader reader, ElementWriter writer, Set<Integer> visited, ClippingPathStack clippingPathStack, Boolean delta,
-                                 List<ElementFeatures> coveredElements, List<ElementFeatures> visibleElements)
-            throws PDFNetException {
-
-        for (Element element = reader.next(); element != null; element = reader.next())
-            switch (element.getType()) {
-
-                case Element.e_image:
-                case Element.e_inline_image:
-                    processImages(element, writer, clippingPathStack, delta, visibleElements);
-                    break;
-
-                case Element.e_text:
-                    processText(element, writer, clippingPathStack, delta, visibleElements);
-                    break;
-
-                case Element.e_path:
-                    processPath(element, writer, clippingPathStack, delta, coveredElements, visibleElements);
-                    break;
-
-                case Element.e_form:
-                    processForm(reader, writer, element, visited, clippingPathStack, delta, coveredElements, visibleElements);
-                    break;
-
-                case Element.e_group_begin:
-                    clippingPathStack.enterNewGState();
-                    writer.writeElement(element);
-                    break;
-
-                case Element.e_group_end:
-                    clippingPathStack.leaveGState();
-                    writer.writeElement(element);
-                    break;
-
-                default:
-                    writer.writeElement(element);
-            }
-    }
-
-    private void processImages(Element imageElement, ElementWriter writer, ClippingPathStack clippingPathStack, Boolean delta, List<ElementFeatures> visibleElements)
-            throws PDFNetException {
-
-        Rect rect = imageElement.getBBox();
-
-        if (rect == null) {
-            return;
-        }
-
-        boolean inClippingPath = clippingPathStack.almostIntersects(rect.getX1(), rect.getY1(), rect.getWidth(), rect.getHeight());
-
-        if (!delta && inClippingPath) {
-            visibleElements.add(ElementFeatures.extractFeatures(imageElement));
-            writer.writeElement(imageElement);
-        }
-
-        if (delta && !inClippingPath) {
-            writer.writeElement(imageElement);
-        }
-    }
-
-    private void processText(Element textElement, ElementWriter writer, ClippingPathStack clippingPathStack,
-                             Boolean delta, List<ElementFeatures> visibleElements)
-            throws PDFNetException {
-
-        Rect rect = textElement.getBBox();
-
-        if (rect == null) {
-            writer.writeElement(textElement);
-            return;
-        }
-
-        GState gState = textElement.getGState();
-
-        boolean inClippingPath = clippingPathStack.almostIntersects(rect.getX1(), rect.getY1(), rect.getWidth(), rect.getHeight());
-
-        boolean isTextVisible = isTextRenderedVisibly(gState);
-
-        if (inClippingPath && isTextVisible) {
-            visibleElements.add(ElementFeatures.extractFeatures(textElement));
-        }
-        if (!delta) {
-            if (inClippingPath && isTextVisible) {
-                writer.writeElement(textElement);
-            } else if (textElement.hasTextMatrix()) {
-                /*
-                PDFTron Element with type "text" refers to a Tj command. If a Tm command is just above it in the pdf file, PDFTron will join the two commands and treat them as one Element.
-                hasTextMatrix() checks for this case specifically. Also, Tm changes the position for a whole BT/ET segment, possibly containing multiple Tj commands.
-                Therefore, the position of a following Tj is affected by not writing the first Element.
-                This is why, we write only the Tm command:
-                */
-                writer.writeGStateChanges(textElement);
-            }
-        } else {
-            if (!inClippingPath) {
-                gState.setFillColorSpace(ColorSpace.createDeviceRGB());
-                gState.setFillColor(new ColorPt(1, 0, 0));
-                writer.writeElement(textElement);
-            }
-            if (!isTextVisible) {
-                gState.setFillColorSpace(ColorSpace.createDeviceRGB());
-                gState.setFillColor(new ColorPt(0, 0, 1));
-                gState.setTextRenderMode(GState.e_fill_text);
-                gState.setFillOpacity(1);
-                writer.writeElement(textElement);
-            }
-        }
-    }
-
-    private void processForm(ElementReader reader, ElementWriter writer, Element element, Set<Integer> visited, ClippingPathStack clippingPathStack, Boolean delta,
-                             List<ElementFeatures> coveredElements, List<ElementFeatures> allElements)
-            throws PDFNetException {
-
-        writer.writeElement(element);
-        Obj formObj = element.getXObject();
-
-        if (!visited.contains((int) formObj.getObjNum())) {
-            visited.add((int) formObj.getObjNum());
-            ElementWriter new_writer = new ElementWriter();
-            reader.formBegin();
-            new_writer.begin(formObj);
-
-            reader.clearChangeList();
-            new_writer.setDefaultGState(reader);
-
-            processElements(reader, new_writer, visited, clippingPathStack, delta, coveredElements, allElements);
-            new_writer.end();
-            reader.end();
-        }
-    }
-
-    private void processPath(Element pathElement, ElementWriter writer, ClippingPathStack clippingPathStack, Boolean delta,
-                             List<ElementFeatures> overlappedElements, List<ElementFeatures> visibleElements)
-            throws PDFNetException {
-
-        GeneralPath linePath = convertToGeneralPath(pathElement.getPathData());
-
-        //transform path to initial user space
-        var ctm = pathElement.getCTM();
-        var affineTransform = getAffineTransform(ctm);
-        linePath.transform(affineTransform);
-
-        var rect = linePath.getBounds2D();
-
-        boolean inClippingPath = clippingPathStack.almostIntersects(rect.getX(), rect.getY(), rect.getWidth(), rect.getHeight());
-
-        if (pathElement.isClippingPath()) {
-            if (pathElement.isClipWindingFill()) {
-                linePath.setWindingRule(GeneralPath.WIND_NON_ZERO);
-            } else {
-                linePath.setWindingRule(GeneralPath.WIND_EVEN_ODD);
-            }
-
-            clippingPathStack.intersectClippingPath(linePath);
-            pathElement.setPathClip(!delta);
-            writer.writeElement(pathElement);
-
-        } else {
-            if (inClippingPath) {
-                if (isFilledAndNonTransparent(pathElement)) {
-                    List<ElementFeatures> currentOverlappedElements = visibleElements.stream()
-                            .filter(features -> almostContains(linePath, features.getBoundingBox()))
-                            .collect(Collectors.toList());
-                    overlappedElements.addAll(currentOverlappedElements);
-                    visibleElements.removeAll(currentOverlappedElements);
-                }
-                visibleElements.add(ElementFeatures.extractFeatures(pathElement));
-                if (!delta) {
-                    writer.writeElement(pathElement);
-                }
-            }
-            if (delta && !inClippingPath) {
-                pathElement.getGState().setFillColorSpace(ColorSpace.createDeviceRGB());
-                pathElement.getGState().setFillColor(new ColorPt(1, 0, 0));
-                pathElement.getGState().setStrokeColorSpace(ColorSpace.createDeviceRGB());
-                pathElement.getGState().setStrokeColor(new ColorPt(1, 0, 0));
-                writer.writeElement(pathElement);
-            }
-        }
-    }
-
-    private static AffineTransform getAffineTransform(Matrix2D ctm) throws PDFNetException {
-        return new AffineTransform(ctm.getA(), ctm.getB(), ctm.getC(), ctm.getD(), ctm.getH(), ctm.getV());
-    }
-
-    private void removeOverlappedElements(Page page, ElementReader reader, ElementWriter writer, Set<Integer> visited, List<ElementFeatures> overlappedElements, boolean delta)
-            throws PDFNetException {
-        reader.begin(page);
-        writer.begin(page, ElementWriter.e_replacement, false, true, page.getResourceDict());
-        if (delta) {
-            overlappedElements.forEach(feature -> drawBBox(writer, feature.getBoundingBox(), "#00FF00"));
-            overlappedElements.clear();
-        }
-        processOverlappedElements(reader, writer, visited, overlappedElements, delta);
-        writer.end();
-        reader.end();
-
-        if (overlappedElements.size() > 0) {
-            log.warn(overlappedElements.size() + " overlapped elements have not been found and removed");
-        }
-    }
-
-    private void processOverlappedElements(ElementReader reader, ElementWriter writer, Set<Integer> visited, List<ElementFeatures> coveredElements, boolean delta)
-            throws PDFNetException {
-        for (Element element = reader.next(); element != null; element = reader.next()) {
-            switch (element.getType()) {
-                case Element.e_form:
-                    processFormOverlappedElements(reader, writer, element, visited, coveredElements, delta);
-                    break;
-                case Element.e_path:
-                case Element.e_image:
-                case Element.e_inline_image:
-                case Element.e_text:
-                    boolean anyMatch = false;
-                    for (ElementFeatures elementToRemove : coveredElements) {
-                        if (elementToRemove.almostMatches(element)) {
-                            coveredElements.remove(elementToRemove);
-                            anyMatch = true;
-                            break;
-                        }
-                    }
-                    if (!anyMatch) {
-                        writer.writeElement(element);
-                    } else if (element.getType() == 3 && element.hasTextMatrix()) {
-                        /*
-                        PDFTron Element with type "text" refers to a Tj command. If a Tm command is just above it in the pdf file, PDFTron will join the two commands and treat them as one Element.
-                        hasTextMatrix() checks for this case specifically. Also, Tm changes the position for a whole BT/ET segment, possibly containing multiple Tj commands.
-                        Therefore, the position of a following Tj is affected by not writing the first Element.
-                        This is why, we write only the Tm command:
-                        */
-                        writer.writeGStateChanges(element);
-                    }
-                    break;
-
-                default:
-                    writer.writeElement(element);
-
-            }
-        }
-    }
-
-    private void processFormOverlappedElements(ElementReader reader, ElementWriter writer, Element element, Set<Integer> visited, List<ElementFeatures> elementsToRemove, boolean delta)
-            throws PDFNetException {
-
-        writer.writeElement(element);
-        Obj formObj = element.getXObject();
-
-        if (!visited.contains((int) formObj.getObjNum())) {
-            visited.add((int) formObj.getObjNum());
-            ElementWriter new_writer = new ElementWriter();
-            reader.formBegin();
-            new_writer.begin(formObj);
-
-            reader.clearChangeList();
-            new_writer.setDefaultGState(reader);
-
-            processOverlappedElements(reader, new_writer, visited, elementsToRemove, delta);
-            new_writer.end();
-            reader.end();
-        }
-    }
-
-    private boolean isTextRenderedVisibly(GState gState) throws PDFNetException {
-        if (gState.getTextRenderMode() == GState.e_invisible_text) return false;
-        if (gState.getTextRenderMode() == GState.e_fill_text && gState.getFillOpacity() == 0) return false;
-        if (gState.getTextRenderMode() == GState.e_stroke_text && gState.getStrokeOpacity() == 0) return false;
-        if (gState.getTextRenderMode() == GState.e_fill_stroke_text && gState.getFillOpacity() == 0 && gState.getStrokeOpacity() == 0) return false;
-        return true;
-    }
-
-    private GeneralPath convertToGeneralPath(PathData pathData) throws PDFNetException {
-        GeneralPath linePath = new GeneralPath();
-
-        double[] dataPoints = pathData.getPoints();
-        byte[] opr = pathData.getOperators();
-
-        double x1;
-        double y1;
-        double x2;
-        double y2;
-        double x3;
-        double y3;
-
-        int data_index = 0;
-        for (int opr_index = 0; opr_index < opr.length; ++opr_index) {
-            switch (opr[opr_index]) {
-                case PathData.e_moveto:
-                    x1 = dataPoints[data_index];
-                    ++data_index;
-                    y1 = dataPoints[data_index];
-                    ++data_index;
-
-                    linePath.moveTo(x1, y1);
-                    break;
-                case PathData.e_lineto:
-                    x1 = dataPoints[data_index];
-                    ++data_index;
-                    y1 = dataPoints[data_index];
-                    ++data_index;
-
-                    linePath.lineTo(x1, y1);
-                    break;
-                case PathData.e_cubicto:
-                    x1 = dataPoints[data_index];
-                    ++data_index;
-                    y1 = dataPoints[data_index];
-                    ++data_index;
-                    x2 = dataPoints[data_index];
-                    ++data_index;
-                    y2 = dataPoints[data_index];
-                    ++data_index;
-                    x3 = dataPoints[data_index];
-                    ++data_index;
-                    y3 = dataPoints[data_index];
-                    ++data_index;
-
-                    linePath.curveTo(x1, y1, x2, y2, x3, y3);
-                    break;
-                case PathData.e_rect:
-                    x1 = dataPoints[data_index];
-                    ++data_index;
-                    y1 = dataPoints[data_index];
-                    ++data_index;
-                    double w = dataPoints[data_index];
-                    ++data_index;
-                    double h = dataPoints[data_index];
-                    ++data_index;
-                    x2 = x1 + w;
-                    y2 = y1;
-                    x3 = x2;
-                    y3 = y1 + h;
-                    double x4 = x1;
-                    double y4 = y3;
-
-                    linePath.moveTo(x1, y1);
-                    linePath.lineTo(x2, y2);
-                    linePath.lineTo(x3, y3);
-                    linePath.lineTo(x4, y4);
-                    break;
-                case PathData.e_closepath:
-                    linePath.closePath();
-                    break;
-                default:
-                    throw new PDFNetException("Invalid Element Type", 0, "", "", "");
-            }
-        }
-        return linePath;
-    }
-
-    private boolean almostContains(Shape outer, Rectangle2D inner) {
-        double tolerance = 1e-3;
-
-        double x_with_tolerance = inner.getX() >= 0 ? inner.getX() + tolerance : inner.getX() - tolerance;
-        double y_with_tolerance = inner.getY() >= 0 ? inner.getY() + tolerance : inner.getY() - tolerance;
-        double height_with_tolerance = inner.getHeight() - (2 * tolerance);
-        double width_with_tolerance = inner.getWidth() - (2 * tolerance);
-        Rectangle2D innerRect = new Rectangle2D.Double(x_with_tolerance, y_with_tolerance, width_with_tolerance, height_with_tolerance);
-
-        return outer.contains(innerRect);
-    }
-
-    private boolean isFilledAndNonTransparent(Element element) throws PDFNetException {
-        return element.isFilled() && element.getGState().getFillOpacity() == 1;
-    }
-
-
-    private void debugSave(byte[] pdfFile) {
-        String fileLocation = "/tmp/delta.pdf";
-        try (var f_out = FileUtils.openOutputStream(new File(fileLocation))) {
-            f_out.write(pdfFile);
-        } catch (IOException e) {
-            throw new RuntimeException("File location: " + fileLocation + "could not be openend, no file will be saved");
-        }
-    }
-
-    @SneakyThrows
-    private void drawBBox(ElementWriter writer, Rectangle2D r, String hexcolor) {
-        ColorPt colorPt = new ColorPt(Integer.valueOf(hexcolor.substring(1, 3), 16) / 255d,
-                Integer.valueOf(hexcolor.substring(3, 5), 16) / 255d,
-                Integer.valueOf(hexcolor.substring(5, 7), 16) / 255d);
-        ElementBuilder eb = new ElementBuilder();
-        Element rect = eb.createRect(r.getX(), r.getY(), r.getWidth(), r.getHeight());
-        rect.setPathStroke(true);
-        rect.getGState().setStrokeColorSpace(ColorSpace.createDeviceRGB());
-        rect.getGState().setStrokeColor(colorPt);
-        writer.writePlacedElement(rect);
-    }
-}
@@ -1,29 +1,31 @@
 package com.iqser.red.service.ocr.v1.server.service;

-import com.fasterxml.jackson.databind.ObjectMapper;
-import com.iqser.red.service.ocr.v1.api.model.OCRStatusUpdateResponse;
-import com.iqser.red.service.ocr.v1.server.configuration.MessagingConfiguration;
-import com.iqser.red.service.ocr.v1.server.model.ImagePosition;
-import com.iqser.red.service.ocr.v1.server.model.image.ImageServiceResponse;
-import com.iqser.red.service.ocr.v1.server.settings.OcrServiceSettings;
-import com.iqser.red.service.persistence.service.v1.api.utils.SuppressFBWarnings;
-import com.iqser.red.service.redaction.v1.model.Point;
-import com.iqser.red.service.redaction.v1.model.Rectangle;
-import com.pdftron.pdf.*;
-import com.pdftron.sdf.SDFDoc;
-import io.micrometer.core.annotation.Timed;
-import lombok.RequiredArgsConstructor;
-import lombok.SneakyThrows;
-import lombok.extern.slf4j.Slf4j;
-import org.apache.commons.io.IOUtils;
-import org.springframework.amqp.rabbit.core.RabbitTemplate;
-import org.springframework.stereotype.Service;
-
 import java.io.ByteArrayInputStream;
 import java.io.ByteArrayOutputStream;
 import java.io.InputStream;
-import java.util.*;
+import java.io.OutputStream;
+import java.util.Map;

+import org.springframework.amqp.rabbit.core.RabbitTemplate;
+import org.springframework.stereotype.Service;
+
+import com.fasterxml.jackson.databind.ObjectMapper;
+import com.iqser.red.service.ocr.v1.api.model.OCRStatusUpdateResponse;
+import com.iqser.red.service.ocr.v1.server.configuration.MessagingConfiguration;
+import com.iqser.red.service.ocr.v1.server.settings.OcrServiceSettings;
+import com.pdftron.common.PDFNetException;
+import com.pdftron.pdf.OCRModule;
+import com.pdftron.pdf.OCROptions;
+import com.pdftron.pdf.Optimizer;
+import com.pdftron.pdf.PDFDoc;
+import com.pdftron.pdf.Page;
+import com.pdftron.pdf.RectCollection;
+import com.pdftron.sdf.SDFDoc;
+
+import io.micrometer.core.annotation.Timed;
+import lombok.RequiredArgsConstructor;
+import lombok.SneakyThrows;
+import lombok.extern.slf4j.Slf4j;

@Slf4j
@Service
@@ -39,139 +41,100 @@ public class OCRService {

    private final ObjectMapper objectMapper;

-    private final InvisibleElementService invisibleElementService;
+    private final InvisibleElementRemovalService invisibleElementRemovalService;

-    @Timed("redactmanager_PDFTron-ocrDocument")
+    private final ImagePositionRetrievalService imagePositionRetrievalService;
+
+
+    /**
+     * First loads the PDF Document from storage.
+     * Then removes all invisible Elements from the PDF, check InvisibleElementRemovalService for details.
+     * Then gets Image Position Information, check ImagePositionRetrievalService for details.
+     * Then runs OCR page by page, exclusively on pages which have images on them. It does so, by creating a new PDFDoc and inserting a single page at a time.
+     * This is because PDFTron OCROptions overlays all regions where OCR should not be run with white images. It does not check for empty pages.
+     * For Documents with many pages but few Images this results in major performance improvements.
+     * It then re-adds the OCRed Pages to the original document and saves it.
+     *
+     * @param dossierId The dossier id
+     * @param fileId    The file id
+     * @param out       OutputStream to write the file to
+     */
    @SneakyThrows
-    public InputStream ocrDocument(String dossierId, String fileId) {
+    @Timed("redactmanager_runOcrOnDocument")
+    public void runOcrOnDocument(String dossierId, String fileId, OutputStream out) {

-        var fileStream = fileStorageService.getOriginalFileAsStream(dossierId, fileId);
-
-        var imageServiceResponse = fileStorageService.getImageServiceResponse(dossierId, fileId);
-
-        var fileBytes = IOUtils.toByteArray(fileStream);
-
-        var ocrBytes = ocr(fileBytes, fileId, imageServiceResponse);
-
-        return new ByteArrayInputStream(ocrBytes);
-
-    }
-
-
-    @SuppressFBWarnings("REC_CATCH_EXCEPTION")
-    private byte[] ocr(byte[] file, String fileId, ImageServiceResponse imageServiceResponse) {
-
-        PDFDoc pdfDoc = null;
-
-        var fileWithoutInvisibleText = invisibleElementService.removeInvisibleTextOrPathElements(file, false);
-
-        try (ByteArrayOutputStream out = new ByteArrayOutputStream()) {
-            pdfDoc = new PDFDoc(fileWithoutInvisibleText);
-
-            Map<Integer, List<ImagePosition>> pages = new HashMap<>();
-
-            // TODO take logic to ignore small and combine images from image-service.
-            // TODO Then replace logic so ocr-service is independent from image-service.
-            imageServiceResponse.getData()
-                    .forEach(imageMetadata -> pages.computeIfAbsent(imageMetadata.getPosition().getPageNumber(), x -> new ArrayList<>())
-                            .add(new ImagePosition(new Rectangle(new Point(imageMetadata.getPosition().getX1(), imageMetadata.getPosition().getY1()),
-                                    imageMetadata.getGeometry().getWidth(),
-                                    imageMetadata.getGeometry().getHeight(),
-                                    imageMetadata.getPosition().getPageNumber()), imageMetadata.isAlpha())));
-
-            Map<Integer, PDFDoc> pdfDocMap = Collections.synchronizedMap(new HashMap<>());
-
-            rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_STATUS_UPDATE_RESPONSE_QUEUE,
-                    objectMapper.writeValueAsString(OCRStatusUpdateResponse.builder().fileId(fileId).numberOfPagesToOCR(pages.keySet().size()).build()));
-
-            ocrPages(pdfDoc, fileId, pages, pdfDocMap);
-
-            for (var entry : pdfDocMap.entrySet()) {
-
-                var ocrDoc = entry.getValue();
-                var page = entry.getKey();
-
-                Page ocrPage = ocrDoc.getPageIterator(1).next();
-                pdfDoc.pageInsert(pdfDoc.getPageIterator(page), ocrPage);
-                pdfDoc.pageRemove(pdfDoc.getPageIterator(page + 1));
-
-                ocrDoc.close();
+        try (ByteArrayOutputStream transferOutputStream = new ByteArrayOutputStream()) {
+            try (InputStream fileStream = fileStorageService.getOriginalFileAsStream(dossierId, fileId)) {
+                invisibleElementRemovalService.removeInvisibleElements(fileStream, transferOutputStream, false);
            }
-            Optimizer.optimize(pdfDoc);
-            pdfDoc.save(out, SDFDoc.SaveMode.LINEARIZED, null);
-            pdfDoc.close();
-
-            rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_STATUS_UPDATE_RESPONSE_QUEUE,
-                    objectMapper.writeValueAsString(OCRStatusUpdateResponse.builder()
-                            .fileId(fileId)
-                            .numberOfPagesToOCR(pages.keySet().size())
-                            .numberOfOCRedPages(pages.keySet().size())
-                            .ocrFinished(true)
-                            .build()));
-
-            return out.toByteArray();
-        } catch (Exception e) {
-            throw new RuntimeException(e);
-        } finally {
-            if (pdfDoc != null) {
-                try {
-                    pdfDoc.close();
-                } catch (Exception e) {
-                    log.debug("Failed to close document", e);
-                }
+            try (InputStream transferInputStream = new ByteArrayInputStream(transferOutputStream.toByteArray())) {
+                runOcr(transferInputStream, out, fileId);
            }
        }
    }


    @SneakyThrows
-    private void ocrPages(PDFDoc pdfDoc, String fileId, Map<Integer, List<ImagePosition>> pages, Map<Integer, PDFDoc> pdfDocMap) {
+    private void runOcr(InputStream fileStream, OutputStream out, String fileId) {

-        int numberOfOCRedPages = 0;
-        for (var pageEntry : pages.entrySet()) {
+        PDFDoc pdfDoc = new PDFDoc(fileStream);

+        Map<Integer, RectCollection> pageIdToRectCollection = imagePositionRetrievalService.getImagePositionPerPage(pdfDoc, true);
+
+        int numProcessedPages = 0;
+        for (Integer pageId : pageIdToRectCollection.keySet()) {
            try {
-                RectCollection rectCollection = new RectCollection();
-
-                var page = pageEntry.getKey();
-
-                Page pdfPage = pdfDoc.getPageIterator(page).next();
-
-                pdfPage.setMediaBox(pdfPage.getCropBox());
-
-                for (ImagePosition imagePosition : pageEntry.getValue()) {
-                    Rectangle rectangle = imagePosition.getRectangle();
-
-                    // Warning coordinate system is different in this call macOs/Linux
-                    double y = -rectangle.getTopLeft().getY() + pdfPage.getCropBox().getY2() - rectangle.getHeight();
-                    rectCollection.addRect(rectangle.getTopLeft().getX(), y, rectangle.getTopLeft().getX() + rectangle.getWidth(), y + rectangle.getHeight());
-                }
-
-                PDFDoc ocrDoc = new PDFDoc();
-                ocrDoc.pagePushBack(pdfPage);
-                pdfDocMap.put(pageEntry.getKey(), ocrDoc);
-
+                // optimization: creating a new document is faster than reusing the same and adding/removing pages one by one
                OCROptions options = new OCROptions();
-                options.addTextZonesForPage(rectCollection, 1);
+                PDFDoc ocrPageDoc = new PDFDoc();
+                // optimization: only scanning pages that contain images
+                Page pdfPage = pdfDoc.getPage(pageId);
+                // optimization: this line ensures the ocr text is placed correctly by PDFTron
+                pdfPage.setMediaBox(pdfPage.getCropBox());
+                ocrPageDoc.pagePushBack(pdfPage);
+                options.addTextZonesForPage(pageIdToRectCollection.get(pageId), 1);
                options.addLang(ENGLISH);
                options.addDPI(settings.getOcrDPI());
-                OCRModule.processPDF(ocrDoc, options);

-                rectCollection.clear();
+                OCRModule.processPDF(ocrPageDoc, options);
+                ++numProcessedPages;

-            } catch (Exception e) {
-                log.warn("Failed to process PDF page {} - {}", pageEntry.getKey(), e);
+                StringBuilder zonesString = new StringBuilder();
+                for (int j = 0; j < pageIdToRectCollection.get(pageId).getNumRects(); ++j) {
+                    var r = pageIdToRectCollection.get(pageId).getRectAt(j);
+                    zonesString.append(String.format("[lower left (%.1f|%.1f) upper right (%.1f|%.1f)]", r.getX1(), r.getY1(), r.getX2(), r.getY2()));
+                }
+                log.info("{}/{} Page {} done, OCR regions {}", numProcessedPages, pageIdToRectCollection.size(), pageId, zonesString);
+
+                // re-adding OCR pages
+                Page ocrPage = ocrPageDoc.getPage(1);
+                pdfDoc.pageInsert(pdfDoc.getPageIterator(pageId), ocrPage);
+                pdfDoc.pageRemove(pdfDoc.getPageIterator(pageId + 1));
+                ocrPageDoc.close();
+
+                rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_STATUS_UPDATE_RESPONSE_QUEUE,
+                        objectMapper.writeValueAsString(OCRStatusUpdateResponse.builder()
+                                .fileId(fileId)
+                                .numberOfPagesToOCR(pageIdToRectCollection.size())
+                                .numberOfOCRedPages(numProcessedPages)
+                                .build()));
+
+            } catch (PDFNetException e) {
+                log.error("failed to process page {}", pageId);
+                throw new RuntimeException(e);
            }
-
-            rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_STATUS_UPDATE_RESPONSE_QUEUE,
-                    objectMapper.writeValueAsString(OCRStatusUpdateResponse.builder()
-                            .fileId(fileId)
-                            .numberOfPagesToOCR(pages.keySet().size())
-                            .numberOfOCRedPages(++numberOfOCRedPages)
-                            .build()));
-
-            log.warn("Done page {}", pageEntry);
-
        }
+
+        rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_STATUS_UPDATE_RESPONSE_QUEUE,
+                objectMapper.writeValueAsString(OCRStatusUpdateResponse.builder()
+                        .fileId(fileId)
+                        .numberOfPagesToOCR(pageIdToRectCollection.size())
+                        .numberOfOCRedPages(numProcessedPages)
+                        .ocrFinished(true)
+                        .build()));
+
+        Optimizer.optimize(pdfDoc);
+        pdfDoc.save(out, SDFDoc.SaveMode.LINEARIZED, null);
    }
+
 }
@@ -1,5 +1,9 @@
 package com.iqser.red.service.ocr.v1.server.service;

+import java.io.ByteArrayInputStream;
+import java.io.ByteArrayOutputStream;
+import java.io.IOException;
+
 import org.springframework.amqp.AmqpRejectAndDontRequeueException;
 import org.springframework.amqp.rabbit.annotation.RabbitHandler;
 import org.springframework.amqp.rabbit.annotation.RabbitListener;
@@ -8,9 +12,9 @@ import org.springframework.stereotype.Service;

 import com.fasterxml.jackson.core.JsonProcessingException;
 import com.fasterxml.jackson.databind.ObjectMapper;
+import com.iqser.red.service.ocr.v1.api.model.DocumentRequest;
 import com.iqser.red.service.ocr.v1.server.client.FileStatusProcessingUpdateClient;
 import com.iqser.red.service.ocr.v1.server.configuration.MessagingConfiguration;
-import com.iqser.red.service.ocr.v1.api.model.DocumentRequest;

 import feign.FeignException;
 import lombok.RequiredArgsConstructor;
@@ -44,9 +48,14 @@ public class OcrMessageReceiver {
            fileStorageService.storeUntouchedFile(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), originalFile);
        }

-        var ocrResult = ocrService.ocrDocument(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId());
+        try (var out = new ByteArrayOutputStream()) {
+            ocrService.runOcrOnDocument(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), out);
+
+            fileStorageService.storeOriginalFile(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), new ByteArrayInputStream(out.toByteArray()));
+        } catch (IOException e) {
+            throw new RuntimeException(e);
+        }

-        fileStorageService.storeOriginalFile(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), ocrResult);

        long end = System.currentTimeMillis();
        log.info("Successfully processed ocr for file with dossierId {} and fileId {}, took {}", ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), end - start);
@@ -35,7 +35,7 @@ management:
    metrics.enabled: ${monitoring.enabled:false}
    prometheus.enabled: ${monitoring.enabled:false}
    health.enabled: true
-  endpoints.web.exposure.include: prometheus, health
+  endpoints.web.exposure.include: prometheus, health, metrics
  metrics.export.prometheus.enabled: ${monitoring.enabled:false}


@@ -46,4 +46,4 @@ storage:
  endpoint: 'https://s3.amazonaws.com'
  backend: 's3'

-pdftron.license: ${PDFTRON_LICENSE}
+pdftron.license: ${PDFTRON_LICENSE}
@@ -1,64 +0,0 @@
-package com.iqser.red.service.ocr.v1.server;
-
-import com.iqser.red.service.ocr.v1.server.service.InvisibleElementService;
-import lombok.SneakyThrows;
-import org.apache.commons.io.FileUtils;
-import org.junit.jupiter.api.Test;
-import org.junit.jupiter.api.extension.ExtendWith;
-import org.springframework.amqp.rabbit.core.RabbitTemplate;
-import org.springframework.beans.factory.annotation.Autowired;
-import org.springframework.boot.test.context.SpringBootTest;
-import org.springframework.boot.test.mock.mockito.MockBean;
-import org.springframework.context.annotation.Import;
-import org.springframework.core.io.ClassPathResource;
-import org.springframework.test.context.junit.jupiter.SpringExtension;
-
-import java.io.File;
-import java.io.IOException;
-import java.nio.file.Files;
-
-import static com.iqser.red.service.ocr.v1.server.utils.OsUtils.getTemporaryDirectory;
-
-@ExtendWith(SpringExtension.class)
-@SpringBootTest(classes = Application.class, webEnvironment = SpringBootTest.WebEnvironment.RANDOM_PORT //
-        , properties = {"pdftron.ocrmodule.path=/YourOCRModulePath"})
-@Import(OcrServiceIntegrationTest.TestConfiguration.class)
-public class InvisibleElementServiceTest {
-
-    @Autowired
-    private InvisibleElementService invisibleElementService;
-
-    @MockBean
-    protected RabbitTemplate rabbitTemplate;
-
-    @Test
-    @SneakyThrows
-    public void testRemoveInvisibleText() {
-        String fileName = "InvisibleText";
-
-        ClassPathResource pdfFileResource = new ClassPathResource("files/" + fileName + ".pdf");
-
-        var initialFileBytes = Files.readAllBytes(pdfFileResource.getFile().toPath());
-
-        var fileWithoutInvisibleText = invisibleElementService.removeInvisibleTextOrPathElements(initialFileBytes, false);
-        var deltaFile = invisibleElementService.removeInvisibleTextOrPathElements(initialFileBytes, true);
-
-        String fileWithoutInvisibleTextLocation = getTemporaryDirectory() + "/" + fileName + ".pdf";
-        String deltaFileLocation = getTemporaryDirectory() + "/" + fileName + "_delta.pdf";
-
-        saveToFile(fileWithoutInvisibleTextLocation, fileWithoutInvisibleText);
-        saveToFile(deltaFileLocation, deltaFile);
-
-        System.out.println("File:" + fileWithoutInvisibleTextLocation);
-        System.out.println("File:" + deltaFileLocation);
-    }
-
-    private void saveToFile(String location, byte[] fileBytes) {
-        try (var f_out = FileUtils.openOutputStream(new File(location))) {
-            f_out.write(fileBytes);
-        } catch (IOException e) {
-            throw new RuntimeException("File location: " + location + "could not be openend, no file will be saved");
-        }
-
-    }
-}
@@ -1,16 +1,18 @@
 package com.iqser.red.service.ocr.v1.server;

-import com.fasterxml.jackson.databind.ObjectMapper;
-import com.iqser.red.service.ocr.v1.server.service.FileStorageService;
-import com.iqser.red.service.ocr.v1.server.service.OCRService;
-import com.iqser.red.service.ocr.v1.server.utils.FileSystemBackedStorageService;
-import com.iqser.red.service.persistence.service.v1.api.model.dossiertemplate.dossier.file.FileType;
-import com.iqser.red.storage.commons.StorageAutoConfiguration;
-import com.iqser.red.storage.commons.service.StorageService;
-import lombok.SneakyThrows;
-import org.apache.commons.io.FileUtils;
-import org.apache.commons.io.IOUtils;
+import static com.iqser.red.service.ocr.v1.server.utils.OsUtils.getTemporaryDirectory;
+import static org.assertj.core.api.Assertions.assertThat;
+
+import java.io.FileInputStream;
+import java.io.FileOutputStream;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.TimeUnit;
+
+import io.micrometer.prometheus.PrometheusMeterRegistry;
+import io.micrometer.prometheus.PrometheusTimer;
 import org.junit.jupiter.api.AfterEach;
+import org.junit.jupiter.api.BeforeEach;
 import org.junit.jupiter.api.Disabled;
 import org.junit.jupiter.api.Test;
 import org.junit.jupiter.api.extension.ExtendWith;
@@ -27,10 +29,20 @@ import org.springframework.context.annotation.Primary;
 import org.springframework.core.io.ClassPathResource;
 import org.springframework.test.context.junit.jupiter.SpringExtension;

-import java.io.File;
+import com.fasterxml.jackson.databind.ObjectMapper;
+import com.iqser.red.service.ocr.v1.server.service.FileStorageService;
+import com.iqser.red.service.ocr.v1.server.service.OCRService;
+import com.iqser.red.service.ocr.v1.server.utils.FileSystemBackedStorageService;
+import com.iqser.red.service.persistence.service.v1.api.model.dossiertemplate.dossier.file.FileType;
+import com.iqser.red.storage.commons.StorageAutoConfiguration;
+import com.iqser.red.storage.commons.service.StorageService;
+import com.pdftron.pdf.OCRModule;
+import com.pdftron.pdf.PDFDoc;
+import com.pdftron.pdf.Page;
+import com.pdftron.pdf.PageIterator;
+import com.pdftron.pdf.TextExtractor;

-import static com.iqser.red.service.ocr.v1.server.utils.OsUtils.getTemporaryDirectory;
-import static org.assertj.core.api.Assertions.assertThat;
+import lombok.SneakyThrows;

@ExtendWith(SpringExtension.class)
@SpringBootTest(classes = Application.class, webEnvironment = SpringBootTest.WebEnvironment.RANDOM_PORT //
@@ -53,55 +65,119 @@ public class OcrServiceIntegrationTest {
    @Autowired
    private OCRService ocrService;

+    @Autowired
+    private PrometheusMeterRegistry registry;

-    @Test
-    @Disabled // OCRModule is not available on build server. If you want to run the test set the property at the top.
+
+    @BeforeEach
    @SneakyThrows
-    public void testOCR() {
+    @Disabled // OCRModule is not available on build server. If you want to run the test set the property at the top.
+    public void assertOCRModuleIsLoaded() {

-        String fileName = "Watermark";
-
-        ClassPathResource imageInfoResource = new ClassPathResource("files/" + fileName + ".IMAGE_INFO.json");
-        ClassPathResource pdfFileResource = new ClassPathResource("files/" + fileName + ".pdf");
-
-        var originId = FileStorageService.getStorageId("dossier", "file", FileType.ORIGIN);
-        storageService.storeObject(originId, pdfFileResource.getInputStream());
-
-        var imageId = FileStorageService.getStorageId("dossier", "file", FileType.IMAGE_INFO);
-        storageService.storeObject(imageId, imageInfoResource.getInputStream());
-
-        var response = ocrService.ocrDocument("dossier", "file");
-
-        var out = FileUtils.openOutputStream(new File(getTemporaryDirectory() + "/" + fileName + ".pdf"));
-        IOUtils.copy(response, out);
-
-        System.out.println("File:" + getTemporaryDirectory() + "/" + fileName + ".pdf");
+        assert OCRModule.isModuleAvailable();
    }


    @Test
    @Disabled // OCRModule is not available on build server. If you want to run the test set the property at the top.
+    public void testOCRMetrics(){
+        testOCR("Watermark");
+        testOCR("Watermark");
+        testOCR("Watermark");
+
+        var ocrOnDocumentMeter = registry.getMeters().stream()
+                .filter(m -> m.getId().getName().equalsIgnoreCase("redactmanager_runOcrOnDocument")).findAny();
+        assertThat(ocrOnDocumentMeter.isPresent()).isTrue();
+        PrometheusTimer timer = (PrometheusTimer) ocrOnDocumentMeter.get();
+        assertThat(timer.count()).isEqualTo(3);
+        assertThat(timer.mean(TimeUnit.SECONDS)).isGreaterThan(0.1);
+    }
+
+    @Test
+    @Disabled // OCRModule is not available on build server. If you want to run the test set the property at the top.
+    public void testOcr() {
+        // check visually for most of the images containing text, the resulting text is kind of nonsense, just ensure it is there
+        String text = testOCR("StitchedImagesMultiPage");
+    }
+
+
+    @Test
+    @Disabled // OCRModule is not available on build server. If you want to run the test set the property at the top.
+    public void testManyRotatedImages() {
+        // check visually for most of the images containing text, the resulting text is kind of nonsense, just ensure it is there
+        String text = testOCR("manyRotatedImages");
+        assertThat(text).contains("Michela", "Gregori", "DVM", "PhD", "Pathologist");
+    }
+
+
+    @Test
+    @Disabled // OCRModule is not available on build server. If you want to run the test set the property at the top.
+    public void testMergeImages() {
+        // check visually for most of the images containing text, the resulting text is kind of nonsense, just ensure it is there
+        String text = testOCR("merge_images");
+        assertThat(text).contains("Bodyweight change of dams with live young - group mean values",
+                "Control",
+                "mg/g day",
+                "10 mg/kg/day",
+                "20 mg/kg/",
+                "Days",
+                "50",
+                "-200",
+                "—250",
+                "150",
+                "200",
+                "250",
+                "—150");
+    }
+
+
+    @Test
+    @Disabled // OCRModule is not available on build server. If you want to run the test set the property at the top.
+    public void testOCRWatermark() {
+
+        assertThat(testOCR("Watermark")).contains("syngenta");
+    }
+
+
+    @Test
+    @Disabled // OCRModule is not available on build server. If you want to run the test set the property at the top.
+    public void testOCRInvisibleText() {
+
+        String text = testOCR("InvisibleText");
+        assertThat(text).contains("Michela", "Gregori", "DVM", "PhD", "Pathologist", "SIGNATURE PAGE");
+        assertThat(text).doesNotContain("COMPLETION DATE:", "LABORATORY PROJECT ID:", "AUTHOR(S):", "Substance");
+    }
+
+
    @SneakyThrows
-    public void testRemoveInvisibleText() {
+    private String testOCR(String fileName) {

-        String fileName = "ocr/OCR Docs/MK244 - Fitness of Analytical Method - Physical-Chemical Pro";
-//        String fileName = "InvisiblePathElements";
-
-        ClassPathResource imageInfoResource = new ClassPathResource("files/InvisibleText.IMAGE_INFO.json");
        ClassPathResource pdfFileResource = new ClassPathResource("files/" + fileName + ".pdf");

        var originId = FileStorageService.getStorageId("dossier", "file", FileType.ORIGIN);
-        storageService.storeObject(originId, pdfFileResource.getInputStream());
+        try (var fileStream = pdfFileResource.getInputStream()) {
+            storageService.storeObject(originId, fileStream);
+        }

-        var imageId = FileStorageService.getStorageId("dossier", "file", FileType.IMAGE_INFO);
-        storageService.storeObject(imageId, imageInfoResource.getInputStream());
+        try (var out = new FileOutputStream(getTemporaryDirectory() + "/" + fileName + ".pdf")) {
+            ocrService.runOcrOnDocument("dossier", "file", out);
+        }

-        var response = ocrService.ocrDocument("dossier", "file");
-
-        var out = FileUtils.openOutputStream(new File(getTemporaryDirectory() + "/" + fileName + ".pdf"));
-        IOUtils.copy(response, out);
+        TextExtractor extractor = new TextExtractor();
+        List<String> texts = new ArrayList<>();
+        PDFDoc pdfDoc;

+        try (var fileStream = new FileInputStream(getTemporaryDirectory() + "/" + fileName + ".pdf")) {
+            pdfDoc = new PDFDoc(fileStream);
+        }
+        PageIterator iterator = pdfDoc.getPageIterator();
+        while (iterator.hasNext()) {
+            Page page = iterator.next();
+            extractor.begin(page);
+            texts.add(extractor.getAsText());
+        }
        System.out.println("File:" + getTemporaryDirectory() + "/" + fileName + ".pdf");
+        return String.join("\n", texts);
    }


@@ -0,0 +1,228 @@
+package com.iqser.red.service.ocr.v1.server.service;
+
+import static com.iqser.red.service.ocr.v1.server.utils.OsUtils.getTemporaryDirectory;
+import static org.assertj.core.api.Assertions.assertThat;
+
+import java.io.FileInputStream;
+import java.io.FileOutputStream;
+import java.io.IOException;
+import java.io.InputStream;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map;
+import java.util.stream.Collectors;
+
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.extension.ExtendWith;
+import org.springframework.amqp.rabbit.core.RabbitTemplate;
+import org.springframework.beans.factory.annotation.Autowired;
+import org.springframework.boot.test.context.SpringBootTest;
+import org.springframework.boot.test.mock.mockito.MockBean;
+import org.springframework.context.annotation.Import;
+import org.springframework.core.io.ClassPathResource;
+import org.springframework.test.context.junit.jupiter.SpringExtension;
+
+import com.iqser.red.service.ocr.v1.server.Application;
+import com.iqser.red.service.ocr.v1.server.OcrServiceIntegrationTest;
+import com.pdftron.common.PDFNetException;
+import com.pdftron.pdf.ColorPt;
+import com.pdftron.pdf.ColorSpace;
+import com.pdftron.pdf.Element;
+import com.pdftron.pdf.ElementBuilder;
+import com.pdftron.pdf.ElementWriter;
+import com.pdftron.pdf.PDFDoc;
+import com.pdftron.pdf.Page;
+import com.pdftron.pdf.Rect;
+import com.pdftron.pdf.RectCollection;
+import com.pdftron.sdf.SDFDoc;
+
+import lombok.SneakyThrows;
+
+@ExtendWith(SpringExtension.class)
+@SpringBootTest(classes = Application.class, webEnvironment = SpringBootTest.WebEnvironment.RANDOM_PORT)
+@Import(OcrServiceIntegrationTest.TestConfiguration.class)
+class ImagePositionRetrievalServiceTest {
+
+    @Autowired
+    private ImagePositionRetrievalService imagePositionRetrievalService;
+
+    @MockBean
+    protected RabbitTemplate rabbitTemplate;
+
+
+    @Test
+    @SneakyThrows
+    public void testImagePositionRetrievalForRotateTestFileWithImages() {
+
+        String fileName = "RotateTestFileWithImages";
+
+        List<int[]> allRectCoords = testImagePositionDetection(fileName);
+
+        assertThat(allRectCoords).contains(new int[]{48, 572, 295, 721},
+                new int[]{54, 279, 301, 428},
+                new int[]{360, 173, 509, 419},
+                new int[]{362, 522, 511, 768},
+                new int[]{459, 354, 608, 600},
+                new int[]{145, 404, 392, 553},
+                new int[]{151, 111, 398, 260},
+                new int[]{457, 5, 606, 251},
+                new int[]{395, 480, 545, 726},
+                new int[]{393, 130, 542, 377},
+                new int[]{88, 236, 334, 386},
+                new int[]{82, 530, 328, 679},
+                new int[]{465, 11, 614, 257},
+                new int[]{159, 117, 406, 266},
+                new int[]{467, 360, 617, 607},
+                new int[]{153, 410, 400, 559});
+    }
+
+
+    @Test
+    @SneakyThrows
+    public void testImagePositionRetrievalForRotateTestFileWithImagesExtremeCropbox() {
+
+        String fileName = "RotateTestFileWithImagesExtremeCropbox";
+
+        List<int[]> allRectCoords = testImagePositionDetection(fileName);
+        assertThat(allRectCoords).contains(new int[]{48, 572, 295, 721},
+                new int[]{362, 522, 511, 768},
+                new int[]{360, 173, 509, 419},
+                new int[]{54, 279, 301, 428},
+                new int[]{145, 192, 392, 341},
+                new int[]{459, 142, 608, 388},
+                new int[]{457, -207, 606, 39},
+                new int[]{151, -101, 398, 48},
+                new int[]{-30, 238, 216, 387},
+                new int[]{283, 188, 433, 434},
+                new int[]{281, -162, 430, 85},
+                new int[]{-24, -56, 222, 94},
+                new int[]{-39, 410, 208, 559},
+                new int[]{275, 360, 425, 607},
+                new int[]{273, 11, 422, 257},
+                new int[]{-33, 117, 214, 266});
+    }
+
+
+    @Test
+    @SneakyThrows
+    public void testMergeImages() {
+
+        String fileName = "merge_images";
+        List<int[]> allRectCoords = testImagePositionDetection(fileName);
+        assertThat(allRectCoords).contains(new int[]{90, 284, 398, 770});
+    }
+
+
+    @Test
+    @SneakyThrows
+    public void testStitchedImagesMultiPage() {
+
+        String fileName = "StitchedImagesMultiPage";
+        List<int[]> allRectCoords = testImagePositionDetection(fileName);
+        assertThat(allRectCoords.size()).isEqualTo(48);
+    }
+
+
+    private List<int[]> testImagePositionDetection(String fileName) throws IOException, PDFNetException {
+
+        InputStream fileStream = new FileInputStream(new ClassPathResource("files/" + fileName + ".pdf").getFile().getAbsolutePath());
+        PDFDoc pdfDoc = new PDFDoc(fileStream);
+
+        Map<Integer, RectCollection> pageIdToRectCollection = imagePositionRetrievalService.getImagePositionPerPage(pdfDoc, false);
+
+        ElementWriter writer = new ElementWriter();
+        pageIdToRectCollection.forEach((pageId, rectCollection) -> {
+            try {
+                writer.begin(pdfDoc.getPage(pageId));
+                drawRectCollection(writer, rectCollection, pdfDoc.getPage(pageId));
+                drawGrid(writer, pdfDoc.getPage(pageId));
+                writer.end();
+                StringBuilder zonesString = new StringBuilder();
+                for (int j = 0; j < rectCollection.getNumRects(); ++j) {
+                    var r = rectCollection.getRectAt(j);
+                    zonesString.append(String.format("%d:[lower left (%.1f|%.1f) upper right (%.1f|%.1f)]; ", j, r.getX1(), r.getY1(), r.getX2(), r.getY2()));
+                }
+                System.out.printf("Page %d: rotation: %d OCR regions %s \n", pageId, pdfDoc.getPage(pageId).getRotation(), zonesString);
+            } catch (PDFNetException e) {
+                throw new RuntimeException(e);
+            }
+        });
+
+        // Check visually for red Rectangles to match images in the saved pdf file
+        try (var out = new FileOutputStream(getTemporaryDirectory() + "/" + fileName + "_IMAGE_BBOX.pdf")) {
+            out.write(pdfDoc.save(SDFDoc.SaveMode.LINEARIZED, null));
+        }
+        System.out.println("result file: " + getTemporaryDirectory() + "/" + fileName + "_IMAGE_BBOX.pdf");
+        fileStream.close();
+        // round all coords to nearest int to account for inconsistencies with the calculation of the bounding box
+        return pageIdToRectCollection.values().stream().map(this::toRoundedCoordinateArrayList).flatMap(List::stream).collect(Collectors.toList());
+    }
+
+
+    @SneakyThrows
+    private List<int[]> toRoundedCoordinateArrayList(RectCollection rectCollection) {
+
+        List<int[]> coords = new ArrayList<>(rectCollection.getNumRects());
+        for (int i = 0; i < rectCollection.getNumRects(); ++i) {
+            var r = rectCollection.getRectAt(i);
+            coords.add(new int[]{(int) Math.round(r.getX1()), (int) Math.round(r.getY1()), (int) Math.round(r.getX2()), (int) Math.round(r.getY2())});
+        }
+        return coords;
+    }
+
+
+    @SneakyThrows
+    private void drawGrid(ElementWriter writer, Page page) {
+
+        ElementBuilder eb = new ElementBuilder();
+        double dX = 15;
+        double dY = 15;
+        int nRows = (int) (page.getPageHeight() / dY) + 1;
+        int nCols = (int) (page.getPageWidth() / dX) + 1;
+        for (int row = 0; row < nRows; ++row) {
+            for (int col = 0; col < nCols; ++col) {
+                Element cell = eb.createRect(col * dX, row * dY, dX, dY);
+                cell.setPathStroke(true);
+                cell.getGState().setLineWidth(1);
+                cell.getGState().setStrokeOpacity(0.1);
+                cell.getGState().setStrokeColorSpace(ColorSpace.createDeviceRGB());
+                if (row == 0 && col == 0) {
+                    cell.getGState().setStrokeColor(new ColorPt(0, 0, 1));
+                    cell.setPathFill(true);
+                    cell.getGState().setFillOpacity(0.8);
+                    cell.getGState().setFillColorSpace(ColorSpace.createDeviceRGB());
+                    cell.getGState().setFillColor(new ColorPt(0, 0, 1));
+                } else {
+                    cell.setPathFill(false);
+                    cell.getGState().setStrokeColor(new ColorPt(0.1, 0.1, 0.1));
+                }
+                writer.writePlacedElement(cell);
+            }
+        }
+    }
+
+
+    @SneakyThrows
+    public static void drawRectCollection(ElementWriter writer, RectCollection rectCollection, Page page) {
+
+        ColorPt colorPt = new ColorPt(1, 0, 0);
+        ElementBuilder eb = new ElementBuilder();
+        for (int i = 0; i < rectCollection.getNumRects(); ++i) {
+            Rect r = rectCollection.getRectAt(i);
+            Element rect = eb.createRect(r.getX1(), r.getY1(), r.getWidth(), r.getHeight());
+
+            rect.setPathStroke(true);
+            rect.getGState().setLineWidth(5);
+            rect.getGState().setStrokeColorSpace(ColorSpace.createDeviceRGB());
+            rect.getGState().setStrokeColor(colorPt);
+
+            rect.setPathFill(true);
+            rect.getGState().setFillColorSpace(ColorSpace.createDeviceRGB());
+            rect.getGState().setFillColor(colorPt);
+            rect.getGState().setFillOpacity(0.5);
+
+            writer.writePlacedElement(rect);
+        }
+    }
+
+}
@@ -0,0 +1,73 @@
+package com.iqser.red.service.ocr.v1.server.service;
+
+import static com.iqser.red.service.ocr.v1.server.utils.OsUtils.getTemporaryDirectory;
+import static org.assertj.core.api.AssertionsForClassTypes.assertThat;
+
+import java.io.FileInputStream;
+import java.io.FileOutputStream;
+
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.extension.ExtendWith;
+import org.springframework.amqp.rabbit.core.RabbitTemplate;
+import org.springframework.beans.factory.annotation.Autowired;
+import org.springframework.boot.test.context.SpringBootTest;
+import org.springframework.boot.test.mock.mockito.MockBean;
+import org.springframework.context.annotation.Import;
+import org.springframework.core.io.ClassPathResource;
+import org.springframework.test.context.junit.jupiter.SpringExtension;
+
+import com.iqser.red.service.ocr.v1.server.Application;
+import com.iqser.red.service.ocr.v1.server.OcrServiceIntegrationTest;
+import com.pdftron.pdf.PDFDoc;
+import com.pdftron.pdf.Page;
+import com.pdftron.pdf.PageIterator;
+import com.pdftron.pdf.TextExtractor;
+
+import lombok.SneakyThrows;
+
+@ExtendWith(SpringExtension.class)
+@SpringBootTest(classes = Application.class, webEnvironment = SpringBootTest.WebEnvironment.RANDOM_PORT)
+@Import(OcrServiceIntegrationTest.TestConfiguration.class)
+public class InvisibleElementRemovalServiceTest {
+
+    @Autowired
+    private InvisibleElementRemovalService invisibleElementRemovalService;
+
+    @MockBean
+    protected RabbitTemplate rabbitTemplate;
+
+
+    @Test
+    @SneakyThrows
+    public void testRemoveInvisibleText() {
+
+        String fileName = "InvisibleText";
+
+        ClassPathResource pdfFileResource = new ClassPathResource("files/" + fileName + ".pdf");
+
+        try (var initialFileStream = pdfFileResource.getInputStream(); var out = new FileOutputStream(getTemporaryDirectory() + "/" + fileName + ".pdf")) {
+            invisibleElementRemovalService.removeInvisibleElements(initialFileStream, out,false);
+        }
+
+        try (var initialFileStream = pdfFileResource.getInputStream(); var out = new FileOutputStream(getTemporaryDirectory() + "/" + fileName + "_delta.pdf")) {
+            invisibleElementRemovalService.removeInvisibleElements(initialFileStream, out,true);
+        }
+
+        System.out.println("Output File without invisible elements: files/" + fileName + ".pdf");
+        System.out.println("Output Delta File without invisible elements: files/" + fileName + "_delta.pdf");
+        TextExtractor extractor = new TextExtractor();
+
+        PDFDoc pdfDoc;
+        try(var fileStream = new FileInputStream(getTemporaryDirectory() + "/" + fileName + ".pdf")) {
+            pdfDoc = new PDFDoc(fileStream);
+        }
+
+        PageIterator iterator = pdfDoc.getPageIterator();
+        while (iterator.hasNext()) {
+            Page page = iterator.next();
+            extractor.begin(page);
+            String[] text = extractor.getAsText().split("\n");
+            assertThat(text).contains("APPENDIX 16 Pathology Report", "Amendment 1", "Page 255 of 260");
+        }
+    }
+}
@@ -5,4 +5,13 @@ spring:
    allow-circular-references: true # FIXME


-pdftron.license: demo:1650351709282:7bd235e003000000004ec28a6743e1163a085e2115de2536ab6e2cfe5a
+pdftron.license: demo:1650351709282:7bd235e003000000004ec28a6743e1163a085e2115de2536ab6e2cfe5a
+
+
+management:
+  endpoint:
+    metrics.enabled: true
+    prometheus.enabled: true
+    health.enabled: true
+  endpoints.web.exposure.include: prometheus, health, metrics
+  metrics.export.prometheus.enabled: true
@@ -1 +0,0 @@
-{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "917b9d9c9f548f85fef3679db45ff46c", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": [{"classification": {"label": "signature", "probabilities": {"signature": 0.9927, "logo": 0.0038, "other": 0.0034, "formula": 0.0}}, "representation": "FFF2CF0F7C74FFC1070830FFF", "position": {"x1": -7, "x2": 603, "y1": 0, "y2": 852, "pageNumber": 1}, "geometry": {"width": 610, "height": 852}, "alpha": false, "filters": {"geometry": {"imageSize": {"quotient": 1.0096, "tooLarge": true, "tooSmall": false}, "imageFormat": {"quotient": 0.716, "tooTall": false, "tooWide": false}}, "probability": {"unconfident": false}, "allPassed": false}}], "dataCV": []}
@@ -1 +0,0 @@
-{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "917b9d9c9f548f85fef3679db45ff46c", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": []}
@@ -1 +0,0 @@
-{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "917b9d9c9f548f85fef3679db45ff46c", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": []}
@@ -1 +0,0 @@
-{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "917b9d9c9f548f85fef3679db45ff46c", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": []}
@@ -1 +0,0 @@
-{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "917b9d9c9f548f85fef3679db45ff46c", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": []}
@@ -1 +0,0 @@
-{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "d7f1e0e37cba4e28ebdf894a79d3bd67", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": [{"classification": {"label": "signature", "probabilities": {"signature": 0.9872, "logo": 0.0064, "other": 0.0063, "formula": 0.0001}}, "representation": "FFFCF10608F6F89747BFFC301", "position": {"x1": -9, "x2": 584, "y1": 9, "y2": 849, "pageNumber": 1}, "geometry": {"width": 593, "height": 840}, "alpha": false, "filters": {"geometry": {"imageSize": {"quotient": 0.9992, "tooLarge": true, "tooSmall": false}, "imageFormat": {"quotient": 0.706, "tooTall": false, "tooWide": false}}, "probability": {"unconfident": false}, "allPassed": false}}], "dataCV": []}
@@ -1 +0,0 @@
-{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "917b9d9c9f548f85fef3679db45ff46c", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": []}
@@ -1 +0,0 @@
-{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "917b9d9c9f548f85fef3679db45ff46c", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": []}
@@ -1 +0,0 @@
-{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "32b19ec38896f5105c09041def470c90", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": [{"classification": {"label": "logo", "probabilities": {"logo": 0.9999, "signature": 0.0001, "formula": 0.0, "other": 0.0}}, "representation": "307EF8F6E9833CE9D7AF9EFFF", "position": {"x1": 26, "x2": 586, "y1": -2, "y2": 794, "pageNumber": 1}, "geometry": {"width": 560, "height": 796}, "alpha": false, "filters": {"geometry": {"imageSize": {"quotient": 0.959, "tooLarge": true, "tooSmall": false}, "imageFormat": {"quotient": 0.7035, "tooTall": false, "tooWide": false}}, "probability": {"unconfident": false}, "allPassed": false}}, {"classification": {"label": "logo", "probabilities": {"logo": 1.0, "formula": 0.0, "other": 0.0, "signature": 0.0}}, "representation": "FFF7FFD2000000018F3FFEFFF", "position": {"x1": 90, "x2": 210, "y1": 676, "y2": 720, "pageNumber": 1}, "geometry": {"width": 120, "height": 44}, "alpha": false, "filters": {"geometry": {"imageSize": {"quotient": 0.1044, "tooLarge": false, "tooSmall": false}, "imageFormat": {"quotient": 2.7273, "tooTall": false, "tooWide": false}}, "probability": {"unconfident": false}, "allPassed": true}}], "dataCV": []}
@@ -1,235 +0,0 @@
-
-0 TL
-q
-    q
-        150 0 m
-        400 0 l
-        400 300 l
-        150 300 l
-        h
-        58 50 m
-        58 200 l
-        570 200 l
-        570 50 l
-        F
-        W
-        n
-        BT
-        -0.011 Tc
-        0 Tw
-        100 Tz
-        /C2_0 12 Tf
-        0 Tr
-        0 Ts
-        485.52 66.48 Td
-        (\0003\000D\000J\000H\000\003) Tj
-        0 Tc
-        26.4 0 Td
-        (\000\024) Tj
-        0.24 Tc
-        8.88 0 Td
-        (\000R\000I) Tj
-        0 Tc
-        9.84 0 Td
-        [(\000\003)-10(\000\030)] TJ
-        0.078 Tc
-        -440.88 -27.6 Td
-        (\0006\000W\000X\000G\000\\) Tj
-        -0.007 Tc
-        30.96 0 Td
-        (\0001\000X\000P\000E\000H\000U\000\035\000\003\000\003) Tj
-        -0.014 Tc
-        48.72 0 Td
-        (\0007\0000\0005\000\023\000\024\000\027\000\025) Tj
-        ET
-        q
-            154.32 0 0 76.32 385.68 643.68 cm
-            /Im0 Do
-        Q
-        BT
-        0.013 Tc
-        /C2_1 12 Tf
-        288.96 601.44 Td
-        (\0006\000X\000E\000V\000W\000D\000Q\000F\000H) Tj
-        0 Tc
-        -168.24 -27.6 Td
-        [(\0006\000<\0001\000\030\000\027\000\031\000\026\000\026\000\023\000\003)-10(\000\261)] TJ
-        ET
-        q
-            192.96 584.88 317.52 -13.92  re
-            W
-            n
-            BT
-            -0.006 Tc
-            192.96 573.84 Td
-            (\0007\000R\000[\000L\000F\000L\000W\000\\\000\003\0006\000W\000X\000G\000\\\000\003\000E\000\\\000\003\000'\000H\000U\000P\000D\000O\000\003\000$\000G\000P\000L\000Q\000L\000V\000W\000U\000D\000W\000L\000R\000Q\000\003\000W\000R\000\003\000+\000D\000Q\000\003\000:\000L\000V\000W\000D\000U\000\003\0005\000D\000W\000V\000\003) Tj
-            ET
-        Q
-        BT
-        -0.024 Tc
-        284.4 560.16 Td
-        (\000I\000R\000U\000\003\000\027\000\003\000:\000H\000H\000N\000V) Tj
-        -0.011 Tc
-        -14.4 -27.6 Td
-        (\0003\000D\000W\000K\000R\000O\000R\000J\000\\\000\003\0005\000H\000S\000R\000U\000W) Tj
-        ET
-        117.12 630 0.72 -3.6  re
-        f*
-        117.12 630 2.16 -0.72  re
-        f*
-        118.56 628.56 0.72 -2.16  re
-        f*
-        118.56 628.56 0.72 -0.72  re
-        f*
-        119.28 630 390.96 -0.72  re
-        f*
-        119.28 628.56 390.96 -0.72  re
-        f*
-        511.68 630 0.72 -3.6  re
-        f*
-        510.24 630 2.16 -0.72  re
-        f*
-        510.24 628.56 0.72 -2.16  re
-        f*
-        510.24 628.56 0.72 -0.72  re
-        f*
-        118.56 626.4 0.72 -111.84  re
-        f*
-        117.12 626.4 0.72 -111.84  re
-        f*
-        117.12 514.56 0.72 -2.16  re
-        f*
-        117.12 513.12 2.16 -0.72  re
-        f*
-        118.56 514.56 0.72 -0.72  re
-        f*
-        118.56 514.56 0.72 -0.72  re
-        f*
-        119.28 513.12 390.96 -0.72  re
-        f*
-        119.28 514.56 390.96 -0.72  re
-        f*
-        511.68 626.4 0.72 -111.84  re
-        f*
-        510.24 626.4 0.72 -111.84  re
-        f*
-        511.68 514.56 0.72 -2.16  re
-        f*
-        510.24 513.12 2.16 -0.72  re
-        f*
-        510.24 514.56 0.72 -0.72  re
-        f*
-        510.24 514.56 0.72 -0.72  re
-        f*
-        BT
-        0.011 Tc
-        89.76 266.64 Td
-        (\000$\0008\0007\000+\0002\0005\000\013\0006\000\f\000\035) Tj
-        -0.012 Tc
-        /C2_0 12 Tf
-        184.8 0.24 Td
-        (\0000\000L\000F\000K\000H\000O\000D\000\003\000*\000U\000H\000J\000R\000U\000L\000\003) Tj
-        -0.037 Tc
-        82.56 0 Td
-        (\000'\0009\0000\000\003\0003\000K\000'\000\003) Tj
-        0.101 Tc
-        55.2 0 Td
-        (\0003\000D\000W\000K\000R\000O) Tj
-        -0.02 Tc
-        30.72 0 Td
-        (\000R\000J\000L\000V\000W) Tj
-        -0.01 Tc
-        /C2_1 12 Tf
-        -353.28 -27.84 Td
-        (\000&\0002\0000\0003\000/\000\(\0007\000,\0002\0001\000\003\000'\000$\0007\000\(\000\035) Tj
-        -0.009 Tc
-        /C2_0 12 Tf
-        184.8 0.24 Td
-        (\000\023\000\030\000\003\0000\000D\000U\000F\000K\000\003\000\025\000\023\000\024\000\033) Tj
-        /C2_1 12 Tf
-        -184.8 -27.84 Td
-        (\000/\000$\000%\0002\0005\000$\0007\0002\0005\000<\000\003\0003\0005\0002\000-\000\(\000&\0007\000\003\000,) Tj
-        -0.026 Tc
-        152.4 0 Td
-        (\000'\000\035) Tj
-        0.009 Tc
-        /C2_0 12 Tf
-        32.4 0.24 Td
-        (\0005\000H\000S\000R\000U\000W\000\003\0001\000X\000P\000E) Tj
-        0.021 Tc
-        65.52 0 Td
-        (\000H\000U\000\035\000\003) Tj
-        -0.014 Tc
-        15.84 0 Td
-        (\0007\0000\0005\000\023\000\024\000\027\000\025) Tj
-        0.078 Tc
-        -81.36 -13.68 Td
-        (\0006\000W\000X\000G\000\\) Tj
-        -0.011 Tc
-        27.84 0 Td
-        (\000\003\0001\000X\000P\000E\000H\000U\000\035) Tj
-        -0.014 Tc
-        48.72 0 Td
-        (\0007\0000\0005\000\023\000\024\000\027\000\025) Tj
-        -0.008 Tc
-        -76.56 -13.92 Td
-        (\0007\000D\000V\000N\000\003\0001\000X\000P\000E\000H\000U\000\035\000\003) Tj
-        -0.019 Tc
-        72 0 Td
-        (\0007\000.\000\023\000\025\000\024\000\023\000\024\000\031\000\033) Tj
-        0 Tc
-        /C2_1 12 Tf
-        -256.8 -27.84 Td
-        [(\0009\0002\000/\0008\0000\000\(\000\003)-10(\000\024)] TJ
-        -0.052 Tc
-        66.24 0 Td
-        (\0002\000\)\000\003) Tj
-        0 Tc
-        19.68 0 Td
-        (\000\024) Tj
-        -0.01 Tc
-        8.88 0 Td
-        (\0002\000\)\000\003\0006\0007\0008\000'\000<) Tj
-        ET
-        146.88 155.04 6 -1.2  re
-        f*
-        175.68 155.04 5.76 -1.2  re
-        f*
-        BT
-        -0.053 Tc
-        89.76 142.56 Td
-        (\0003\000$\000*\000\(\000\003) Tj
-        0 Tc
-        36.24 0 Td
-        (\000\024) Tj
-        -0.052 Tc
-        9.12 0 Td
-        (\0002\000\)\000\003) Tj
-        0 Tc
-        19.68 0 Td
-        (\000\030) Tj
-        ET
-        126 141.36 6 -1.2  re
-        f*
-        154.8 141.36 6 -1.2  re
-        f*
-        /Artifact <</Subtype/Watermark/Type/Pagination>> BDC
-        q
-            /G0 gs
-            0.940613 0 0 0.940613 26.0628 0 cm
-            0 0 0 RG
-            0 w
-            /Fm0 Do
-        Q
-        EMC
-        /Artifact <</Contents( Page 250 of 256)/Subtype/Header/Type/Pagination>> BDC
-        q
-            /G0 gs
-            1 0 0 1 458.67 48.412 cm
-            0 0 0 RG
-            0 w
-            /Fm1 Do
-        Q
-        EMC
-    Q
-Q
@@ -0,0 +1,16 @@
+<Configuration>
+
+    <Appenders>
+        <Console name="CONSOLE" target="SYSTEM_OUT">
+            <PatternLayout pattern="%d{HH:mm:ss.SSS} [%t] %-5level %logger{36} - %msg%n"/>
+        </Console>
+    </Appenders>
+
+    <Loggers>
+        <Root level="warn">
+            <AppenderRef ref="CONSOLE"/>
+        </Root>
+        <Logger name="com.iqser" level="info"/>
+    </Loggers>
+
+</Configuration>
@@ -7,7 +7,7 @@
    <parent>
        <groupId>com.iqser.red</groupId>
        <artifactId>platform-dependency</artifactId>
-        <version>1.14.0</version>
+        <version>1.16.0</version>
        <relativePath/>
    </parent>
Author	SHA1	Message	Date
Kilian Schuettler	6ccf3f80fc	RED-6126: performance-test re-enabled overlap detection re-creating helper document for every page instead of reusing and adding/removing pages	2023-02-09 11:22:39 +01:00
Kilian Schuettler	e705f869fd	RED-6126: Performance Tests moved to streams for pdf file transfer disabled overlap detection	2023-02-09 11:09:52 +01:00
Timo Bejan	efaa291e43	Pull request #6 : RED-4609 - added ocr metric, enabled prometheus, added test for metric Merge in RED/ocr-service from RED-4609 to master * commit '7c71d8ad041f839c21ec26023ee8eaef670a4924': RED-4609 - added ocr metric, enabled prometheus, added test for metric	2023-02-09 10:57:37 +01:00
Timo Bejan	7c71d8ad04	RED-4609 - added ocr metric, enabled prometheus, added test for metric	2023-02-08 16:46:51 +02:00
Kilian Schuettler	b0a658213d	Pull request #5 : RED-6126 Merge in RED/ocr-service from RED-6126 to master * commit '00cfe9e44948c153857ad59442dbc9349e1d4555': RED-6126: In the OCRService, OCR Text is not applied to Document reformatted InvisibleElementRemovalService with new Code Style RED-6126: In the OCRService, OCR Text is not applied to Document updated some comments very slight refactor RED-6126: In the OCRService, OCR Text is not applied to Document complete refactor of the OCRService moved image position retrieval to new class instead of image service added new tests for image rotation RED-6126: In the OCRService, OCR Text is not applied to Document removed private configuration RED-6126: In the OCRService, OCR Text is not applied to Document formatted one line RED-6126: In the OCRService, OCR Text is not applied to Document reverted application of OCR Text to Document to old state refactored OCR Service slightly *added meaningful test cases	2023-02-07 13:35:32 +01:00
Kilian Schuettler	00cfe9e449	RED-6126: In the OCRService, OCR Text is not applied to Document *reformatted InvisibleElementRemovalService with new Code Style	2023-02-07 12:52:09 +01:00
Kilian Schuettler	d0d6bf70a4	RED-6126: In the OCRService, OCR Text is not applied to Document updated some comments very slight refactor	2023-02-07 12:09:04 +01:00
Kilian Schuettler	a415224db5	RED-6126: In the OCRService, OCR Text is not applied to Document complete refactor of the OCRService moved image position retrieval to new class instead of image service *added new tests for image rotation	2023-02-07 12:05:24 +01:00
Kilian Schuettler	355887c865	RED-6126: In the OCRService, OCR Text is not applied to Document *removed private configuration	2023-02-03 13:16:56 +01:00
Kilian Schuettler	ab566a11a9	RED-6126: In the OCRService, OCR Text is not applied to Document *formatted one line	2023-02-03 13:03:47 +01:00
Kilian Schuettler	edd044395e	RED-6126: In the OCRService, OCR Text is not applied to Document reverted application of OCR Text to Document to old state refactored OCR Service slightly *added meaningful test cases	2023-02-03 13:01:01 +01:00
Kilian Schuettler	b37ec5afc9	Pull request #4 : RED-6019: Remove hidden text when processing OCR Merge in RED/ocr-service from RED-6019 to master * commit 'a96260f77fd5b546a5d27d84f34861742f13ddff': RED-6019: Remove hidden text when processing OCR moved InvisibleElementRemovalDto to private inner record of InvisibleElementRemovalService added comments for color choices RED-6019: Remove hidden text when processing OCR moved to release version of platform-dependencies restored annotationProcessors RED-6019: Remove hidden text when processing OCR code refactor upgrade to java 17 RED-6019: Remove hidden text when processing OCR handled cases: Text which is transparent or is set to not render Elements outside of clipping path Elements that have been painted over by visible and filled Paths unhandled cases: Elements covered by widely stroked path Elements same color as background Any Text set to clipping with its many interactions with other elements	2023-02-02 13:05:03 +01:00
Kilian Schuettler	a96260f77f	RED-6019: Remove hidden text when processing OCR moved InvisibleElementRemovalDto to private inner record of InvisibleElementRemovalService added comments for color choices	2023-02-02 13:01:58 +01:00
Kilian Schuettler	12fbdbee50	RED-6019: Remove hidden text when processing OCR moved to release version of platform-dependencies restored annotationProcessors	2023-02-02 10:53:19 +01:00
Kilian Schuettler	99a0cb51d0	RED-6019: Remove hidden text when processing OCR code refactor upgrade to java 17	2023-02-02 10:27:01 +01:00
Kilian Schuettler	fd7ec6e7aa	RED-6019: Remove hidden text when processing OCR handled cases: Text which is transparent or is set to not render Elements outside of clipping path Elements that have been painted over by visible and filled Paths unhandled cases: Elements covered by widely stroked path Elements same color as background Any Text set to clipping with its many interactions with other elements	2023-01-30 16:13:51 +01:00
Dominique Eiflaender	265fac8099	Pull request #3 : RED-5911: Reverted to old ocr logic that uses ContentReplacer/TextExtractor to remove text behind images Merge in RED/ocr-service from RED-5911 to master * commit '7a4c5c2f898e83623a66ef29ab9ed696e2057e24': RED-5911: Reverted to old ocr logic that uses ContentReplacer/TextExtractor to remove text behind images	2023-01-17 12:42:10 +01:00
deiflaender	7a4c5c2f89	RED-5911: Reverted to old ocr logic that uses ContentReplacer/TextExtractor to remove text behind images	2023-01-17 12:15:34 +01:00
				`@@ -1 +0,0 @@`
				{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "917b9d9c9f548f85fef3679db45ff46c", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": [{"classification": {"label": "signature", "probabilities": {"signature": 0.9927, "logo": 0.0038, "other": 0.0034, "formula": 0.0}}, "representation": "FFF2CF0F7C74FFC1070830FFF", "position": {"x1": -7, "x2": 603, "y1": 0, "y2": 852, "pageNumber": 1}, "geometry": {"width": 610, "height": 852}, "alpha": false, "filters": {"geometry": {"imageSize": {"quotient": 1.0096, "tooLarge": true, "tooSmall": false}, "imageFormat": {"quotient": 0.716, "tooTall": false, "tooWide": false}}, "probability": {"unconfident": false}, "allPassed": false}}], "dataCV": []}
				`@@ -1 +0,0 @@`
				`{"dossierId": "55547c91-6b0e-4aa6-9009-2e7c4cd90f13", "fileId": "917b9d9c9f548f85fef3679db45ff46c", "targetFileExtension": "ORIGIN.pdf.gz", "responseFileExtension": "IMAGE_INFO.json.gz", "data": []}`