Pull request #19: RED-4875 1
Merge in RED/ocr-service from RED-4875_1 to master * commit '036203c24a1f5eb1588e945eb020a666cac5dba2': RED-4875 - delete commented out classes RED-4875 - set version of pdftron-common-logics to newest (release) RED-4875 - removed duration log RED-4875 - update version of pdftron-commons to newest RED-4875 - update version of pdftron logic commons to newest RED-4875 - set version of common pdftron logics to newest and move PdfTextExtraction to this new repo RED-4875 - update version of pdftron-logic-commons to newest RED-4875 - call logic of new repo pdftron-logic-commons instead of local one
This commit is contained in:
Thomas Beyer
commit
d9e4f79099
@ -23,6 +23,12 @@
|
||||
<groupId>com.iqser.red.commons</groupId>
|
||||
<artifactId>storage-commons</artifactId>
|
||||
</dependency>
|
||||
<dependency>
|
||||
<groupId>com.iqser.red.commons</groupId>
|
||||
<artifactId>pdftron-logic-commons</artifactId>
|
||||
<version>1.1.0</version>
|
||||
</dependency>
|
||||
|
||||
<dependency>
|
||||
<groupId>com.iqser.red.commons</groupId>
|
||||
<artifactId>spring-commons</artifactId>
|
||||
|
||||
@ -10,6 +10,7 @@ import org.springframework.context.annotation.Bean;
|
||||
import org.springframework.context.annotation.Import;
|
||||
import org.springframework.scheduling.annotation.EnableAsync;
|
||||
|
||||
import com.iqser.red.pdftronlogic.commons.InvisibleElementRemovalService;
|
||||
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.server.multitenancy.AsyncConfig;
|
||||
@ -44,4 +45,11 @@ public class Application {
|
||||
return new TimedAspect(registry);
|
||||
}
|
||||
|
||||
|
||||
@Bean
|
||||
public InvisibleElementRemovalService invisibleElementRemovalService() {
|
||||
|
||||
return new InvisibleElementRemovalService();
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -1,68 +0,0 @@
|
||||
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.Area;
|
||||
import java.awt.geom.GeneralPath;
|
||||
import java.awt.geom.Rectangle2D;
|
||||
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 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 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();
|
||||
}
|
||||
|
||||
}
|
||||
@ -1,170 +0,0 @@
|
||||
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.AccessLevel;
|
||||
import lombok.EqualsAndHashCode;
|
||||
import lombok.Getter;
|
||||
import lombok.SneakyThrows;
|
||||
import lombok.experimental.FieldDefaults;
|
||||
import lombok.experimental.SuperBuilder;
|
||||
|
||||
@Getter
|
||||
@SuperBuilder
|
||||
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
|
||||
public class ElementFeatures {
|
||||
|
||||
int elementType;
|
||||
Rectangle2D boundingBox;
|
||||
|
||||
|
||||
public boolean almostMatches(Element element) throws PDFNetException {
|
||||
|
||||
return element.getType() == elementType && //
|
||||
element.getBBox() != null && //
|
||||
rectsAlmostMatch(element.getBBox());
|
||||
}
|
||||
|
||||
|
||||
protected boolean almostEqual(double a, double b) {
|
||||
|
||||
return Math.abs(a - b) < TOLERANCE;
|
||||
}
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
private boolean rectsAlmostMatch(Rect bBox) {
|
||||
// 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)
|
||||
@Getter
|
||||
@SuperBuilder
|
||||
@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 {
|
||||
|
||||
return super.almostMatches(element) && //
|
||||
text.equals(element.getTextString()) && //
|
||||
font == element.getGState().getFont().getType() && //
|
||||
almostEqual(fontsize, element.getGState().getFontSize());
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@EqualsAndHashCode(callSuper = true)
|
||||
@Getter
|
||||
@SuperBuilder
|
||||
@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 {
|
||||
|
||||
return super.almostMatches(element) && //
|
||||
isClippingPath == element.isClippingPath() && //
|
||||
isClipWindingFill == element.isClipWindingFill() && //
|
||||
isStroked == element.isStroked() && //
|
||||
isFilled == element.isFilled() && //
|
||||
isWindingFill == element.isWindingFill();
|
||||
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@EqualsAndHashCode(callSuper = true)
|
||||
@Getter
|
||||
@SuperBuilder
|
||||
@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 {
|
||||
|
||||
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 {
|
||||
|
||||
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());
|
||||
}
|
||||
|
||||
}
|
||||
@ -1,466 +0,0 @@
|
||||
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);
|
||||
}
|
||||
|
||||
try {
|
||||
pdfDoc.save(out, SDFDoc.SaveMode.LINEARIZED, null);
|
||||
} catch (Exception e) {
|
||||
log.error("File could not be saved after invisible element removal");
|
||||
throw new RuntimeException(e);
|
||||
}
|
||||
|
||||
writer.destroy();
|
||||
reader.destroy();
|
||||
pdfDoc.close();
|
||||
}
|
||||
|
||||
|
||||
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();
|
||||
formWriter.destroy();
|
||||
context.reader().end();
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
private void processPath(Element pathElement, ElementWriter writer, InvisibleElementRemovalContext context) throws PDFNetException {
|
||||
|
||||
PathData pathData = pathElement.getPathData();
|
||||
|
||||
if (pathData.getOperators().length == 0 && pathData.getPoints().length == 0) {
|
||||
writer.writeGStateChanges(pathElement);
|
||||
return;
|
||||
}
|
||||
|
||||
GeneralPath linePath = convertToGeneralPath(pathData);
|
||||
|
||||
//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();
|
||||
formWriter.destroy();
|
||||
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);
|
||||
|
||||
colorPt.destroy();
|
||||
eb.destroy();
|
||||
}
|
||||
|
||||
|
||||
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) {
|
||||
|
||||
}
|
||||
|
||||
}
|
||||
@ -14,6 +14,7 @@ import org.springframework.amqp.rabbit.core.RabbitTemplate;
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import com.fasterxml.jackson.databind.ObjectMapper;
|
||||
import com.iqser.red.pdftronlogic.commons.InvisibleElementRemovalService;
|
||||
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;
|
||||
@ -68,14 +69,7 @@ public class OCRService {
|
||||
|
||||
try (ByteArrayOutputStream transferOutputStream = new ByteArrayOutputStream()) {
|
||||
try (InputStream fileStream = fileStorageService.getOriginalFileAsStream(dossierId, fileId)) {
|
||||
long removalStart = System.currentTimeMillis();
|
||||
log.debug("Start invisible element removal for file with dossierId {} and fileId {}", dossierId, fileId);
|
||||
invisibleElementRemovalService.removeInvisibleElements(fileStream, transferOutputStream, false);
|
||||
long removalEnd = System.currentTimeMillis();
|
||||
log.debug("Invisible element removal successful for file with dossierId {} and fileId {}, took {}s",
|
||||
dossierId,
|
||||
fileId,
|
||||
format("%.1f", (removalEnd - removalStart) / 1000.0));
|
||||
}
|
||||
try (InputStream transferInputStream = new ByteArrayInputStream(transferOutputStream.toByteArray())) {
|
||||
long ocrStart = System.currentTimeMillis();
|
||||
|
||||
@ -1,7 +1,7 @@
|
||||
package com.iqser.red.service.ocr.v1.server;
|
||||
|
||||
import static com.iqser.red.pdftronlogic.commons.PdfTextExtraction.extractAllTextFromDocument;
|
||||
import static com.iqser.red.service.ocr.v1.server.utils.OsUtils.getTemporaryDirectory;
|
||||
import static com.iqser.red.service.ocr.v1.server.utils.PdfTextExtraction.extractAllTextFromDocument;
|
||||
import static org.assertj.core.api.Assertions.assertThat;
|
||||
|
||||
import java.io.FileInputStream;
|
||||
|
||||
@ -1,7 +1,7 @@
|
||||
package com.iqser.red.service.ocr.v1.server.service;
|
||||
|
||||
import static com.iqser.red.pdftronlogic.commons.PdfTextExtraction.extractAllTextFromDocument;
|
||||
import static com.iqser.red.service.ocr.v1.server.utils.OsUtils.getTemporaryDirectory;
|
||||
import static com.iqser.red.service.ocr.v1.server.utils.PdfTextExtraction.extractAllTextFromDocument;
|
||||
import static org.assertj.core.api.AssertionsForClassTypes.assertThat;
|
||||
|
||||
import java.io.FileInputStream;
|
||||
@ -9,16 +9,18 @@ import java.io.FileOutputStream;
|
||||
|
||||
import org.junit.jupiter.api.Test;
|
||||
import org.springframework.beans.factory.annotation.Autowired;
|
||||
import org.springframework.context.annotation.Bean;
|
||||
import org.springframework.core.io.ClassPathResource;
|
||||
|
||||
import com.iqser.red.pdftronlogic.commons.InvisibleElementRemovalService;
|
||||
import com.iqser.red.service.ocr.v1.server.AbstractTest;
|
||||
|
||||
import lombok.SneakyThrows;
|
||||
|
||||
public class InvisibleElementRemovalServiceTest extends AbstractTest {
|
||||
|
||||
@Autowired
|
||||
private InvisibleElementRemovalService invisibleElementRemovalService;
|
||||
@Autowired
|
||||
private InvisibleElementRemovalService invisibleElementRemovalService;
|
||||
|
||||
|
||||
@Test
|
||||
@ -44,5 +46,6 @@ public class InvisibleElementRemovalServiceTest extends AbstractTest {
|
||||
String[] text = extractAllTextFromDocument(fileStream).split("\n");
|
||||
assertThat(text).contains("APPENDIX 16 Pathology Report", "Amendment 1", "Page 255 of 260");
|
||||
}
|
||||
|
||||
}
|
||||
}
|
||||
@ -1,35 +0,0 @@
|
||||
package com.iqser.red.service.ocr.v1.server.utils;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.io.InputStream;
|
||||
import java.util.ArrayList;
|
||||
import java.util.List;
|
||||
|
||||
import com.pdftron.common.PDFNetException;
|
||||
import com.pdftron.pdf.PDFDoc;
|
||||
import com.pdftron.pdf.Page;
|
||||
import com.pdftron.pdf.PageIterator;
|
||||
import com.pdftron.pdf.TextExtractor;
|
||||
|
||||
|
||||
public class PdfTextExtraction {
|
||||
|
||||
public static String extractAllTextFromDocument(InputStream fileStream) throws IOException, PDFNetException {
|
||||
|
||||
PDFDoc pdfDoc = new PDFDoc(fileStream);
|
||||
TextExtractor extractor = new TextExtractor();
|
||||
List<String> texts = new ArrayList<>();
|
||||
|
||||
PageIterator iterator = pdfDoc.getPageIterator();
|
||||
while (iterator.hasNext()) {
|
||||
Page page = iterator.next();
|
||||
extractor.begin(page);
|
||||
texts.add(extractor.getAsText());
|
||||
}
|
||||
|
||||
extractor.destroy();
|
||||
pdfDoc.close();
|
||||
return String.join("\n", texts);
|
||||
}
|
||||
|
||||
}
|
||||
Loading…
x
Reference in New Issue
Block a user