Merge branch 'RED-7075' into 'master'

Resolve RED-7075

Closes RED-7075

See merge request redactmanager/commons/pdftron-logic-commons!11

pom.xml

@@ -26,6 +26,12 @@
             <artifactId>slf4j-api</artifactId>
             <scope>provided</scope>
         </dependency>
+        <dependency>
+            <groupId>org.apache.logging.log4j</groupId>
+            <artifactId>log4j-slf4j2-impl</artifactId>
+            <version>2.20.0</version>
+            <scope>test</scope>
+        </dependency>
         <dependency>
             <groupId>com.google.guava</groupId>
             <artifactId>guava</artifactId>

src/main/java/com/iqser/red/pdftronlogic/commons/ElementFeatureFactory.java

@@ -0,0 +1,80 @@
+package com.iqser.red.pdftronlogic.commons;
+
+import com.pdftron.common.PDFNetException;
+import com.pdftron.pdf.Element;
+
+public class ElementFeatureFactory {
+    public static ElementFeatures extractFeatures(Element element) throws PDFNetException {
+
+        return switch (element.getType()) {
+            case Element.e_path -> buildPath(element);
+            case Element.e_text -> buildText(element);
+            case Element.e_image, Element.e_inline_image -> buildImage(element).build();
+            case Element.e_form -> buildForm(element);
+            // 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());
+        };
+    }
+
+    public static ElementFeatures extractFeaturesWithHash(Element element, String hashObject) throws PDFNetException {
+        return buildImage(element)
+                .hashOfImage(hashObject)
+                .build();
+    }
+
+
+    private static ElementFeatures.Form buildForm(Element element) throws PDFNetException {
+
+        return ElementFeatures.Form.builder()
+                .elementType(element.getType())
+                .boundingBox(Converter.toRectangle2D(element.getBBox()))
+                .xObjectType(element.getXObject().getType())
+                .dictOrArrayOrStreamLength(element.getXObject().getType() == 7 ? element.getXObject().getDecodedStream().size() : 0)
+                .build();
+    }
+
+
+    private static ElementFeatures.Image.ImageBuilder<?, ?> buildImage(Element element) throws PDFNetException {
+
+        return ElementFeatures.Image.builder()
+                .elementType(element.getType())
+                .boundingBox(Converter.toRectangle2D(element.getBBox()))
+                .dataSize(element.getImageDataSize())
+                .height(element.getImageHeight())
+                .width(element.getImageWidth())
+                .renderingIntent(element.getImageRenderingIntent())
+                .componentNum(element.getComponentNum())
+                .bitsPerComponent(element.getBitsPerComponent());
+    }
+
+
+    private static ElementFeatures.Text buildText(Element element) throws PDFNetException {
+
+        return ElementFeatures.Text.builder()
+                .elementType(element.getType())
+                .boundingBox(Converter.toRectangle2D(element.getBBox()))
+                .text(element.getTextString())
+                .font(element.getGState().getFont().getType())
+                .fontsize(element.getGState().getFontSize())
+                .build();
+    }
+
+
+    private static ElementFeatures.Path buildPath(Element element) throws PDFNetException {
+
+        return ElementFeatures.Path.builder()
+                .elementType(element.getType())
+                .boundingBox(Converter.toRectangle2D(element.getBBox()))
+                .isClippingPath(element.isClippingPath())
+                .isClipWindingFill(element.isClipWindingFill())
+                .isStroked(element.isStroked())
+                .isFilled(element.isFilled())
+                .isWindingFill(element.isWindingFill())
+                .fillColor(Converter.convertColor(element.getGState().getFillColorSpace(), element.getGState().getFillColor()))
+                .strokeColor(Converter.convertColor(element.getGState().getStrokeColorSpace(), element.getGState().getStrokeColor()))
+                .linePath(Converter.convertToGeneralPathAndTransformToInitialUserSpace(element.getPathData(), element.getCTM()))
+                .build();
+    }
+
+
+}

src/main/java/com/iqser/red/pdftronlogic/commons/ElementFeatures.java

@@ -26,43 +26,6 @@ public class ElementFeatures {
     int elementType;
     Rectangle2D boundingBox;
 
-    public static ElementFeatures extractFeatures(Element element) throws PDFNetException {
-
-        return switch (element.getType()) {
-            case Element.e_path -> Path.builder()
-                    .elementType(element.getType())
-                    .boundingBox(Converter.toRectangle2D(element.getBBox()))
-                    .isClippingPath(element.isClippingPath())
-                    .isClipWindingFill(element.isClipWindingFill())
-                    .isStroked(element.isStroked())
-                    .isFilled(element.isFilled())
-                    .isWindingFill(element.isWindingFill())
-                    .fillColor(Converter.convertColor(element.getGState().getFillColorSpace(), element.getGState().getFillColor()))
-                    .strokeColor(Converter.convertColor(element.getGState().getStrokeColorSpace(), element.getGState().getStrokeColor()))
-                    .linePath(Converter.convertToGeneralPathAndTransformToInitialUserSpace(element.getPathData(), element.getCTM()))
-                    .build();
-            case Element.e_text -> Text.builder()
-                    .elementType(element.getType())
-                    .boundingBox(Converter.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(Converter.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());
-        };
-    }
-
 
     public boolean almostMatches(Element element) throws PDFNetException {
 
@@ -71,6 +34,12 @@ public class ElementFeatures {
                 rectsAlmostMatch(element.getBBox());
     }
 
+    public boolean almostMatches(ElementFeatures elementFeatures){
+        return elementFeatures.getElementType() == elementType &&
+                elementFeatures.getBoundingBox() != null &&
+                rectsAlmostMatch(elementFeatures.getBoundingBox());
+    }
+
 
     protected boolean almostEqual(double a, double b) {
 
@@ -88,12 +57,22 @@ public class ElementFeatures {
                 almostEqual(bBox.getHeight(), boundingBox.getHeight());
     }
 
+    @SneakyThrows
+    private boolean rectsAlmostMatch(Rectangle2D bBox) {
+        // To address the inconsistencies in the calculation of the bounding box we check equality with a tolerance
+
+        return almostEqual(bBox.getX(), boundingBox.getX()) && //
+                almostEqual(bBox.getY(), boundingBox.getY()) && //
+                almostEqual(bBox.getWidth(), boundingBox.getWidth()) && //
+                almostEqual(bBox.getHeight(), boundingBox.getHeight());
+    }
+
 
     @EqualsAndHashCode(callSuper = true)
     @Getter
     @SuperBuilder
     @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
-    private static class Text extends ElementFeatures {
+    public static class Text extends ElementFeatures {
 
         String text;
         int font;
@@ -159,7 +138,7 @@ public class ElementFeatures {
     @Getter
     @SuperBuilder
     @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
-    private static class Image extends ElementFeatures {
+    public static class Image extends ElementFeatures {
 
         int dataSize;
         int height;
@@ -167,7 +146,7 @@ public class ElementFeatures {
         int renderingIntent;
         int componentNum;
         int bitsPerComponent;
-
+        String hashOfImage;
 
         @Override
         public boolean almostMatches(Element element) throws PDFNetException {
@@ -181,6 +160,73 @@ public class ElementFeatures {
                     bitsPerComponent == element.getBitsPerComponent();
         }
 
+        public boolean almostMatches(ElementFeatures elementFeatures){
+            if(elementFeatures.getClass() != this.getClass()){
+                return false;
+            }
+            return super.almostMatches(elementFeatures) &&
+                    this.dataSize == ((Image) elementFeatures).getDataSize() &&
+                    this.height == ((Image) elementFeatures).getHeight() &&
+                    this.width == ((Image) elementFeatures).getWidth() &&
+                    this.renderingIntent == ((Image) elementFeatures).getRenderingIntent() &&
+                    this.componentNum == ((Image) elementFeatures).getComponentNum() &&
+                    this.bitsPerComponent == ((Image) elementFeatures).getBitsPerComponent() &&
+                    calculateHammingDistance(((Image) elementFeatures).getHashOfImage()) <=4;
+        }
+
+        // Helper method to calculate the Hamming distance between two hexadecimal strings
+        private int calculateHammingDistance(String hash2) {
+            int distance = 0;
+            int maxLength = Math.max(this.hashOfImage.length(), hash2.length());
+            for (int i = 0; i < maxLength; i++) {
+                char char1 = i < this.hashOfImage.length() ? this.hashOfImage.charAt(i) : '0';
+                char char2 = i < hash2.length() ? hash2.charAt(i) : '0';
+                if (char1 != char2) {
+                    distance++;
+                }
+            }
+            return distance;
+        }
+
+    }
+
+    @EqualsAndHashCode(callSuper = true)
+    @Getter
+    @SuperBuilder
+    @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
+    public static class Form extends ElementFeatures {
+
+        int xObjectType;
+        long dictOrArrayOrStreamLength;
+
+
+        @Override
+        public boolean almostMatches(Element element) throws PDFNetException {
+            return element.getType() == getElementType() && //
+                    element.getBBox() != null && //
+                    (super.rectsAlmostMatch(element.getBBox()) || almostRotateMatches(element.getBBox().getRectangle())) &&
+                    xObjectType == element.getXObject().getType() &&
+                    dictOrArrayOrStreamLength == element.getXObject().getDecodedStream().size();
+        }
+
+        public boolean almostMatches(ElementFeatures elementFeatures){
+            if(elementFeatures.getClass() != this.getClass()){
+                return false;
+            }
+            return elementFeatures.getElementType() == getElementType() &&
+                    elementFeatures.getBoundingBox() != null &&
+                    (super.rectsAlmostMatch(elementFeatures.getBoundingBox()) || almostRotateMatches(elementFeatures.getBoundingBox().getBounds2D())) &&
+                    xObjectType == ((Form)elementFeatures).getXObjectType() &&
+                    dictOrArrayOrStreamLength == ((Form)elementFeatures).getDictOrArrayOrStreamLength();
+
+        }
+
+
+        private boolean almostRotateMatches(Rectangle2D bBox) {
+            return almostEqual(bBox.getWidth(), getBoundingBox().getHeight()) && //
+                    almostEqual(bBox.getHeight(), getBoundingBox().getWidth());
+        }
+
     }
 
 

src/main/java/com/iqser/red/pdftronlogic/commons/ImageHashFactory.java

@@ -0,0 +1,96 @@
+package com.iqser.red.pdftronlogic.commons;
+
+import java.awt.Color;
+import java.awt.image.BufferedImage;
+import java.io.ByteArrayInputStream;
+
+import javax.imageio.ImageIO;
+
+import com.pdftron.filters.FilterWriter;
+import com.pdftron.filters.MemoryFilter;
+import com.pdftron.pdf.Element;
+
+import lombok.SneakyThrows;
+import lombok.experimental.UtilityClass;
+
+@UtilityClass
+public class ImageHashFactory {
+
+    @SneakyThrows
+    public String calculate(Element element) {
+
+        com.pdftron.pdf.Image image = new com.pdftron.pdf.Image(element.getXObject());
+
+        byte[] imageBytes = getBytesOfImage(image);
+        ByteArrayInputStream byteArrayInputStream = new ByteArrayInputStream(imageBytes);
+        BufferedImage image1 = ImageIO.read(byteArrayInputStream);
+
+        String hash = getSimplePHash(image1);
+
+        return hash;
+
+    }
+
+
+    @SneakyThrows
+    private byte[] getBytesOfImage(com.pdftron.pdf.Image inputImage) {
+        // 0 because the memory filter determines the size
+        var memFilter = new MemoryFilter(0, false);
+        var filterWriter = new FilterWriter(memFilter);
+
+        inputImage.export(filterWriter);
+        filterWriter.flushAll();
+        byte[] res = memFilter.getBuffer();
+
+        memFilter.flushAll();
+        memFilter.destroy();
+        filterWriter.destroy();
+        return res;
+    }
+
+
+    public String getSimplePHash(BufferedImage image) {
+        // Resize the image to a fixed size (e.g., 8x8 pixels)
+        int targetWidth = 8;
+        int targetHeight = 8;
+        BufferedImage resizedImage = new BufferedImage(targetWidth, targetHeight, BufferedImage.TYPE_INT_ARGB);
+        resizedImage.getGraphics().drawImage(image.getScaledInstance(targetWidth, targetHeight, java.awt.Image.SCALE_SMOOTH), 0, 0, targetWidth, targetHeight, null);
+
+        // Convert the image to grayscale
+        BufferedImage grayscaleImage = new BufferedImage(targetWidth, targetHeight, BufferedImage.TYPE_BYTE_GRAY);
+        grayscaleImage.getGraphics().drawImage(resizedImage, 0, 0, null);
+
+        // Calculate the average grayscale pixel value
+        int average = calculateAverage(grayscaleImage);
+
+        // Create a binary hash based on pixel values
+        StringBuilder hashBuilder = new StringBuilder();
+        for (int y = 0; y < targetHeight; y++) {
+            for (int x = 0; x < targetWidth; x++) {
+                int pixelValue = new Color(grayscaleImage.getRGB(x, y)).getRed();
+                if (pixelValue > average) {
+                    hashBuilder.append("1");
+                } else {
+                    hashBuilder.append("0");
+                }
+            }
+        }
+        return hashBuilder.toString();
+    }
+
+
+    // Helper method to calculate the average grayscale pixel value
+    private int calculateAverage(BufferedImage image) {
+
+        int total = 0;
+        int width = image.getWidth();
+        int height = image.getHeight();
+        for (int y = 0; y < height; y++) {
+            for (int x = 0; x < width; x++) {
+                total += new Color(image.getRGB(x, y)).getRed();
+            }
+        }
+        return total / (width * height);
+    }
+
+}

src/main/java/com/iqser/red/pdftronlogic/commons/InvisibleElementRemovalService.java

@@ -167,7 +167,7 @@ public class InvisibleElementRemovalService {
         boolean inClippingPath = context.clippingPathStack().almostIntersects(rect.getX1(), rect.getY1(), rect.getWidth(), rect.getHeight());
 
         if (!context.delta() && inClippingPath) {
-            context.visibleElements().add(ElementFeatures.extractFeatures(imageElement));
+            context.visibleElements().add(ElementFeatureFactory.extractFeatures(imageElement));
         }
 
         if (context.delta() ^ inClippingPath) {
@@ -192,7 +192,7 @@ public class InvisibleElementRemovalService {
         boolean isTextVisible = isTextRenderedVisibly(gState, textBBox, context);
 
         if (inClippingPath && isTextVisible) {
-            context.visibleElements().add(ElementFeatures.extractFeatures(textElement));
+            context.visibleElements().add(ElementFeatureFactory.extractFeatures(textElement));
         }
         if (!context.delta()) {
             if (inClippingPath && isTextVisible) {
@@ -291,7 +291,7 @@ public class InvisibleElementRemovalService {
                     context.overlappedElements().addAll(currentOverlappedElements);
                     context.visibleElements().removeAll(currentOverlappedElements);
                 }
-                context.visibleElements().add(ElementFeatures.extractFeatures(pathElement));
+                context.visibleElements().add(ElementFeatureFactory.extractFeatures(pathElement));
                 if (!context.delta()) {
                     writer.writeElement(pathElement);
                 }

src/main/java/com/iqser/red/pdftronlogic/commons/WatermarkRemovalService.java

@@ -0,0 +1,289 @@
+package com.iqser.red.pdftronlogic.commons;
+
+import java.io.InputStream;
+import java.io.OutputStream;
+import java.util.Collection;
+import java.util.HashMap;
+import java.util.LinkedList;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+import java.util.TreeSet;
+
+import com.pdftron.common.PDFNetException;
+import com.pdftron.pdf.Element;
+import com.pdftron.pdf.ElementReader;
+import com.pdftron.pdf.ElementWriter;
+import com.pdftron.pdf.PDFDoc;
+import com.pdftron.pdf.Page;
+import com.pdftron.pdf.PageIterator;
+import com.pdftron.sdf.SDFDoc;
+
+import lombok.SneakyThrows;
+import lombok.experimental.UtilityClass;
+import lombok.extern.slf4j.Slf4j;
+
+@UtilityClass
+@Slf4j
+public class WatermarkRemovalService {
+
+    final static double AREA_THRESHOLD = 0.6; // multiplied with page area
+    final static double OCCURING_ON_PAGES_THRESHOLD_FACTOR = 0.75; // multiplied with number of pages
+
+    final static int MIN_PAGES_THRESHOLD = 3;
+
+
+    /**
+     * The method remove watermark works only for Documents with size greater than MIN_PAGES_THRESHOLD.
+     * First the possible watermarks (big XObjects or Images) will be detected and then checked if those appear on most pages according to the
+     * OCCURING_ON_PAGES_THRESHOLD_FACTOR by using image hashing for similarity and size and stream size of the xobjects.
+     * If so, these detected and confirmed will not be written to the pdf file.
+     * @param pdfFile PDFFile to remove watermarks
+     * @param out The OutputStream the final file will be written to
+     */
+    @SneakyThrows
+    public void removeWatermarks(InputStream pdfFile, OutputStream out) {
+
+        PDFDoc pdfDoc = new PDFDoc(pdfFile);
+
+        if(pdfDoc.getPageCount() < MIN_PAGES_THRESHOLD){
+            log.info("Document page count {} is below threshold {}", pdfDoc.getPageCount(), MIN_PAGES_THRESHOLD);
+        } else {
+            Map<Long, List<ElementFeatures>> formObjectsForPages = findAllFormObjectsAndImages(pdfDoc);
+
+            List<ElementFeatures> watermarkElementFeatures = filterSameFormObjectsOccuringOnMostPages(formObjectsForPages);
+
+            if (watermarkElementFeatures.size() > 0) {
+                log.info("Watermark found and will be removed!");
+                removeAllWatermarks(pdfDoc, watermarkElementFeatures);
+            } else {
+                log.info("No watermark found!");
+            }
+        }
+
+        try {
+            pdfDoc.save(out, SDFDoc.SaveMode.LINEARIZED, null);
+        } catch (Exception e) {
+            throw new RuntimeException(e);
+        } finally {
+            pdfDoc.close();
+        }
+    }
+
+
+    @SneakyThrows
+    private Map<Long, List<ElementFeatures>> findAllFormObjectsAndImages(PDFDoc pdfDoc) {
+
+        List<ElementFeatures> formObjectsOccuringMoreThanOnceOnAPage = new LinkedList<>();
+        Map<Long, List<ElementFeatures>> formObjectsAndImagesForPages = new HashMap<>();
+        Set<Long> visitedXObjIds = new TreeSet<>();
+
+        ElementReader reader = new ElementReader();
+
+
+
+        for (PageIterator iterator = pdfDoc.getPageIterator(); iterator.hasNext(); ) {
+
+            Page page = iterator.next();
+
+            double minAreaCoveringFromPage = AREA_THRESHOLD * page.getPageHeight() * page.getPageWidth();
+
+            LinkedList<ElementFeatures> elementFeaturesLinkedList = new LinkedList<>();
+
+            reader.begin(page);
+            for (Element element = reader.next(); element != null; element = reader.next()) {
+                processElement(element, visitedXObjIds, elementFeaturesLinkedList, formObjectsOccuringMoreThanOnceOnAPage, minAreaCoveringFromPage);
+            }
+
+            formObjectsAndImagesForPages.put(page.getSDFObj().getObjNum(), elementFeaturesLinkedList);
+        }
+
+        reader.destroy();
+
+        return formObjectsAndImagesForPages;
+    }
+
+
+    private void processElement(Element element,
+                                       Set<Long> visitedXObjIds,
+                                       List<ElementFeatures> elementFeaturesLinkedList,
+                                       List<ElementFeatures> formObjectsOccuringMoreThanOnceOnAPage,
+                                       double minAreaCoveringPage) throws PDFNetException {
+
+        if (element.getBBox() == null) {
+            return;
+        }
+        if (element.getBBox().getHeight() * element.getBBox().getWidth() < minAreaCoveringPage) {
+            return;
+        }
+
+        if (element.getType() == Element.e_form) {
+            processXObject(element, visitedXObjIds, elementFeaturesLinkedList, formObjectsOccuringMoreThanOnceOnAPage, minAreaCoveringPage);
+        } else if (element.getType() == Element.e_image || element.getType() == Element.e_inline_image) {
+            processImages(element, elementFeaturesLinkedList);
+        }
+    }
+
+
+    @SneakyThrows
+    private void processImages(Element element, List<ElementFeatures> elementFeaturesLinkedList) {
+
+        String hashOfImage = ImageHashFactory.calculate(element);
+        ElementFeatures elementFeatures = ElementFeatureFactory.extractFeaturesWithHash(element, hashOfImage);
+        elementFeaturesLinkedList.add(elementFeatures);
+    }
+
+
+    @SneakyThrows
+    private void processXObject(Element element,
+                                       Set<Long> visitedXObjIds,
+                                       List<ElementFeatures> elementFeaturesLinkedList,
+                                       List<ElementFeatures> formObjectsOccuringMoreThanOnceOnAPage,
+                                       double minAreaCoveringPage) {
+
+        if (visitedXObjIds.add(element.getXObject().getObjNum())) {
+            ElementReader xObjectReader = new ElementReader();
+            xObjectReader.begin(element.getXObject());
+            for (Element element1 = xObjectReader.next(); element1 != null; element1 = xObjectReader.next()) {
+                processElement(element1, visitedXObjIds, elementFeaturesLinkedList, formObjectsOccuringMoreThanOnceOnAPage, minAreaCoveringPage);
+            }
+            elementFeaturesLinkedList.add(ElementFeatureFactory.extractFeatures(element));
+            xObjectReader.destroy();
+        } else {
+            elementFeaturesLinkedList.add(ElementFeatureFactory.extractFeatures(element));
+        }
+    }
+
+
+    /*
+        parameter
+     */
+    private List<ElementFeatures> filterSameFormObjectsOccuringOnMostPages(Map<Long, List<ElementFeatures>> formObjectsPerPage) {
+
+        int pageCount = formObjectsPerPage.keySet().size();
+        int minPagesFilter = (int) (OCCURING_ON_PAGES_THRESHOLD_FACTOR * pageCount);
+
+        return formObjectsPerPage.values()
+                .stream()
+                .flatMap(Collection::stream)
+                .filter(elementFeature -> formObjectsPerPage.values()
+                        .stream()
+                        .filter(elementFeaturesOnPage -> elementFeaturesOnPage.stream().anyMatch(elementFeature::almostMatches))
+                        .count() >= minPagesFilter)
+                .toList();
+    }
+
+
+    @SneakyThrows
+    private void removeAllWatermarks(PDFDoc pdfDoc, List<ElementFeatures> watermarksElementFeaturesList) {
+
+        ElementReader reader = new ElementReader();
+        ElementWriter writer = new ElementWriter();
+        Set<Long> visitedXObjIds = new TreeSet<>();
+
+        for (PageIterator iterator = pdfDoc.getPageIterator(); iterator.hasNext(); ) {
+
+            Page page = iterator.next();
+
+            writeAllElementsExceptWatermarks(page, reader, writer, watermarksElementFeaturesList, visitedXObjIds);
+
+        }
+
+        reader.destroy();
+        writer.destroy();
+
+    }
+
+
+    @SneakyThrows
+    private void writeAllElementsExceptWatermarks(Page page,
+                                                         ElementReader reader,
+                                                         ElementWriter writer,
+                                                         List<ElementFeatures> watermarksElementFeaturesList,
+                                                         Set<Long> visitedXObjIds) {
+
+        reader.begin(page);
+        writer.begin(page, ElementWriter.e_replacement, false, true, page.getResourceDict());
+        processElements(page, reader, writer, watermarksElementFeaturesList, visitedXObjIds);
+        writer.end();
+        reader.end();
+    }
+
+
+    private void processElements(Page page,
+                                        ElementReader reader,
+                                        ElementWriter writer,
+                                        List<ElementFeatures> watermarksElementFeaturesList,
+                                        Set<Long> visitedXObjIds) throws PDFNetException {
+
+        double minAreaCoveringFromPage = AREA_THRESHOLD * page.getPageHeight() * page.getPageWidth();
+        for (Element element = reader.next(); element != null; element = reader.next()) {
+
+            switch (element.getType()) {
+                case Element.e_image, Element.e_inline_image -> {
+                    if (element.getBBox() == null) {
+                        continue;
+                    }
+                    if (element.getBBox().getHeight() * element.getBBox().getWidth() < minAreaCoveringFromPage) {
+                        writer.writeElement(element);
+                        continue;
+                    }
+                    removeImages(element, writer, watermarksElementFeaturesList);
+                }
+                case Element.e_form -> processForms(page, element, reader, writer, watermarksElementFeaturesList, visitedXObjIds);
+                default -> writer.writeElement(element);
+            }
+        }
+    }
+
+
+    @SneakyThrows
+    private void removeImages(Element element, ElementWriter writer, List<ElementFeatures> watermarksElementFeaturesList) {
+
+        String hashValueOfImage = ImageHashFactory.calculate(element);
+        ElementFeatures imageFeatures = ElementFeatureFactory.extractFeaturesWithHash(element, hashValueOfImage);
+        for (ElementFeatures elementFeatures : watermarksElementFeaturesList) {
+            if (elementFeatures.almostMatches(imageFeatures)) {
+                return;
+            }
+        }
+
+        writer.writeElement(element);
+    }
+
+
+    private void processForms(Page page,
+                                     Element element,
+                                     ElementReader reader,
+                                     ElementWriter writer,
+                                     List<ElementFeatures> watermarksElementFeaturesList,
+                                     Set<Long> visitedXObjIds) throws PDFNetException {
+
+        for (ElementFeatures elementFeatures : watermarksElementFeaturesList) {
+            if (elementFeatures.almostMatches(element)) {
+                return;
+            }
+        }
+
+        writer.writeElement(element);
+
+        if (!visitedXObjIds.contains(element.getXObject().getObjNum())) {
+            visitedXObjIds.add(element.getXObject().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();
+            reader.formBegin();
+            formWriter.begin(element.getXObject());
+
+            reader.clearChangeList();
+            formWriter.setDefaultGState(reader);
+
+            processElements(page, reader, formWriter, watermarksElementFeaturesList, visitedXObjIds);
+            formWriter.end();
+            formWriter.destroy();
+            reader.end();
+        }
+
+    }
+
+}

src/test/java/com/iqser/red/pdftronlogic/commons/WatermarkRemovalServiceTest.java

@@ -0,0 +1,62 @@
+package com.iqser.red.pdftronlogic.commons;
+
+import static org.junit.jupiter.api.Assertions.*;
+
+import java.io.File;
+import java.io.FileInputStream;
+import java.io.FileOutputStream;
+import java.nio.file.Path;
+import java.util.Locale;
+
+import org.junit.jupiter.api.Test;
+import org.junit.platform.commons.util.StringUtils;
+
+import com.pdftron.pdf.PDFNet;
+
+import lombok.SneakyThrows;
+
+class WatermarkRemovalServiceTest {
+
+    @SneakyThrows
+    @Test
+    void removeWatermarks() {
+
+        PDFNet.addResourceSearchPath("C:/Users/RaphaelArnold/knecon/pdftron/ocrirismodule/Lib");
+        PDFNet.initialize("demo:1650351709282:7bd235e003000000004ec28a6743e1163a085e2115de2536ab6e2cfe5a");
+
+        String filename = "files/PO_82_277_H_2013 (1).pdf";
+
+        String tmpFilename = createTmpFileName(filename, "WATERMARK_REMOVAL");
+        try (var in = this.getClass().getClassLoader().getResourceAsStream(filename); var out = new FileOutputStream(tmpFilename)) {
+
+            {
+                System.out.println(tmpFilename);
+                WatermarkRemovalService.removeWatermarks(in, out);
+            }
+        }
+
+    }
+
+
+    private static boolean isWindows() {
+
+        return System.getProperty("os.name").toLowerCase(Locale.ROOT).contains("windows");
+    }
+
+
+    public static String getTemporaryDirectory() {
+
+        String tmpdir = System.getProperty("java.io.tmpdir");
+        if (isWindows() && StringUtils.isNotBlank(tmpdir)) {
+            return tmpdir;
+        }
+        return "/tmp";
+    }
+
+
+    public static String createTmpFileName(String filename, String suffix) {
+
+        return Path.of(getTemporaryDirectory()).resolve(Path.of(filename).getFileName()).toString().replace(".pdf", "_" + suffix + ".pdf");
+    }
+
+}