com.itextpdf.pdfocr.tesseract4.TesseractHelper Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of pdfocr-tesseract4 Show documentation
Show all versions of pdfocr-tesseract4 Show documentation
pdfOCR-Tesseract4 is an iText add-on for Java to recognize and extract text in scanned documents and images. It can also convert them into fully ISO-compliant PDF or PDF/A-3u files that are accessible, searchable, and suitable for archiving
/*
This file is part of the iText (R) project.
Copyright (c) 1998-2024 Apryse Group NV
Authors: Apryse Software.
This program is offered under a commercial and under the AGPL license.
For commercial licensing, contact us at https://itextpdf.com/sales. For AGPL licensing, see below.
AGPL licensing:
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see > parseHocrFile(
final List inputFiles, final List txtInputFiles,
final Tesseract4OcrEngineProperties tesseract4OcrEngineProperties)
throws IOException {
Map> imageData =
new LinkedHashMap>();
Map unparsedBBoxes = new LinkedHashMap<>();
for (int inputFileIdx = 0; inputFileIdx < inputFiles.size(); inputFileIdx++) {
final File inputFile = inputFiles.get(inputFileIdx);
List txt = null;
if (txtInputFiles != null) {
final File txtInputFile = txtInputFiles.get(inputFileIdx);
txt = Files.readAllLines(txtInputFile.toPath(), StandardCharsets.UTF_8);
}
if (inputFile != null
&& Files.exists(
java.nio.file.Paths
.get(inputFile.getAbsolutePath()))) {
FileInputStream fileInputStream =
new FileInputStream(inputFile.getAbsolutePath());
Document doc = Jsoup.parse(fileInputStream,
java.nio.charset.StandardCharsets.UTF_8.name(),
inputFile.getAbsolutePath());
Elements pages = doc.getElementsByClass(OCR_PAGE);
for (Element page : pages) {
String[] pageNum = page.id().split(PAGE_PREFIX_PATTERN);
int pageNumber = Integer
.parseInt(pageNum[pageNum.length - 1]);
final List textData = getTextData(page,
tesseract4OcrEngineProperties,
txt,
unparsedBBoxes);
if (textData.size() > 0) {
if (imageData.containsKey(pageNumber)) {
pageNumber = Collections.max(imageData.keySet())
+ 1;
}
imageData.put(pageNumber, textData);
}
}
fileInputStream.close();
}
}
for (Node node : unparsedBBoxes.values()) {
LOGGER.warn(MessageFormatUtil.format(
Tesseract4LogMessageConstant.CANNOT_PARSE_NODE_BBOX,
node.toString()
));
}
return imageData;
}
/**
* Get and align (if needed) bbox of the element.
*/
static Rectangle getAlignedBBox(Element object,
TextPositioning textPositioning,
Rectangle pageBbox,
Map unparsedBBoxes) {
final Rectangle box = parseBBox(object, pageBbox, unparsedBBoxes);
if (TextPositioning.BY_WORDS_AND_LINES == textPositioning
|| TextPositioning.BY_WORDS == textPositioning) {
Node line = object.parent();
final Rectangle lineBbox = parseBBox(line, pageBbox, unparsedBBoxes);
if (TextPositioning.BY_WORDS_AND_LINES == textPositioning) {
box.setBbox(box.getLeft(),
lineBbox.getBottom(),
box.getRight(),
lineBbox.getTop());
}
detectAndFixBrokenBBoxes(object, box,
lineBbox, pageBbox, unparsedBBoxes);
}
return box;
}
/**
* Parses element bbox.
*
* @param node element containing bbox
* @param pageBBox element containing parent page bbox
* @param unparsedBBoxes list of element ids with bboxes which could not be parsed
* @return parsed bbox
*/
static Rectangle parseBBox(Node node, Rectangle pageBBox, Map unparsedBBoxes) {
List bbox = new ArrayList<>();
Matcher bboxMatcher = BBOX_PATTERN.matcher(node.attr(TITLE));
if (bboxMatcher.matches()) {
Matcher bboxCoordinateMatcher =
BBOX_COORDINATE_PATTERN
.matcher(bboxMatcher.group());
if (bboxCoordinateMatcher.matches()) {
for (int i = 0; i < BBOX_ARRAY_SIZE; i++) {
String coord = bboxCoordinateMatcher
.group(i + 1);
bbox.add(Float.parseFloat(coord));
}
}
}
if (bbox.size() == 0) {
bbox = Arrays.asList(0f, 0f, 0f, 0f);
String id = node.attr("id");
if (id != null && !unparsedBBoxes.containsKey(id)) {
unparsedBBoxes.put(id, node);
}
}
if (pageBBox == null) {
return new Rectangle(toPoints(bbox.get(LEFT_IDX)),
toPoints(bbox.get(TOP_IDX)),
toPoints(bbox.get(RIGHT_IDX)),
toPoints(bbox.get(BOTTOM_IDX) - bbox.get(TOP_IDX)));
} else {
return new Rectangle(0, 0).setBbox(toPoints(bbox.get(LEFT_IDX)),
pageBBox.getTop() - toPoints(bbox.get(TOP_IDX)),
toPoints(bbox.get(RIGHT_IDX)),
pageBBox.getTop() - toPoints(bbox.get(BOTTOM_IDX)));
}
}
/**
* Sometimes hOCR file contains broke character bboxes which are equal to page bbox.
* This method attempts to detect and fix them.
*/
static void detectAndFixBrokenBBoxes(Element object, Rectangle bbox,
Rectangle lineBbox, Rectangle pageBbox,
Map unparsedBBoxes) {
if (bbox.getLeft() < lineBbox.getLeft()
|| bbox.getLeft() > lineBbox.getRight()) {
if (object.previousElementSibling() == null) {
bbox.setX(lineBbox.getLeft());
} else {
Element sibling = object.previousElementSibling();
final Rectangle siblingBBox = parseBBox(sibling, pageBbox, unparsedBBoxes);
bbox.setX(siblingBBox.getRight());
}
}
if (bbox.getRight() > lineBbox.getRight()
|| bbox.getRight() < lineBbox.getLeft()) {
if (object.nextElementSibling() == null) {
bbox.setBbox(bbox.getLeft(),
bbox.getBottom(),
lineBbox.getRight(),
bbox.getTop());
} else {
Element sibling = object.nextElementSibling();
final Rectangle siblingBBox = parseBBox(sibling, pageBbox, unparsedBBoxes);
bbox.setBbox(bbox.getLeft(),
bbox.getBottom(),
siblingBBox.getLeft(),
bbox.getTop());
}
}
}
/**
* Converts points to pixels.
*/
static float toPixels(float pt) {
return pt / PX_TO_PT;
}
/**
* Converts pixels to points.
*/
static float toPoints(float px) {
return px * PX_TO_PT;
}
/**
* Deletes file using provided path.
*
* @param pathToFile path to the file to be deleted
*/
static void deleteFile(final String pathToFile) {
try {
if (pathToFile != null && !pathToFile.isEmpty()
&& Files.exists(java.nio.file.Paths.get(pathToFile))) {
Files.delete(java.nio.file.Paths.get(pathToFile));
}
} catch (IOException | SecurityException e) {
LOGGER.info(MessageFormatUtil.format(
Tesseract4LogMessageConstant.CANNOT_DELETE_FILE,
pathToFile,
e.getMessage()));
}
}
/**
* Reads from text file to string.
*
* @param txtFile input {@link java.io.File} to be read
* @return result {@link java.lang.String} from provided text file
*/
static String readTxtFile(final File txtFile) {
String content = null;
try {
content = new String(
Files.readAllBytes(txtFile.toPath()),
StandardCharsets.UTF_8);
} catch (IOException e) {
LOGGER.error(MessageFormatUtil.format(
Tesseract4LogMessageConstant.CANNOT_READ_FILE,
txtFile.getAbsolutePath(),
e.getMessage()));
}
return content;
}
/**
* Writes provided {@link java.lang.String} to text file using
* provided path.
*
* @param path path as {@link java.lang.String} to file to be created
* @param data text data in required format as {@link java.lang.String}
*/
static void writeToTextFile(final String path,
final String data) {
try (Writer writer = new OutputStreamWriter(new FileOutputStream(path),
StandardCharsets.UTF_8)) {
writer.write(data);
} catch (IOException e) {
throw new PdfOcrInputTesseract4Exception(
PdfOcrTesseract4ExceptionMessageConstant.CANNOT_WRITE_TO_FILE, e);
}
}
/**
* Runs given command.
*
* @param execPath path to the executable
* @param paramsList {@link java.util.List} of command line arguments
* @throws PdfOcrTesseract4Exception if provided command failed
*/
static void runCommand(final String execPath,
final List paramsList) throws PdfOcrTesseract4Exception {
runCommand(execPath, paramsList, null);
}
/**
* Runs given command from the specific working directory.
*
* @param execPath path to the executable
* @param paramsList {@link java.util.List} of command line arguments
* @param workingDirPath path to the working directory
* @throws PdfOcrTesseract4Exception if provided command failed
*/
static void runCommand(final String execPath,
final List paramsList,
final String workingDirPath) throws PdfOcrTesseract4Exception {
try {
String params = String.join(" ", paramsList);
boolean cmdSucceeded = SystemUtil
.runProcessAndWait(execPath, params, workingDirPath);
if (!cmdSucceeded) {
LOGGER.error(MessageFormatUtil
.format(Tesseract4LogMessageConstant.COMMAND_FAILED,
execPath + " " + params));
throw new PdfOcrTesseract4Exception(
PdfOcrTesseract4ExceptionMessageConstant
.TESSERACT_FAILED);
}
} catch (Exception e) {
LOGGER.error(MessageFormatUtil
.format(Tesseract4LogMessageConstant.COMMAND_FAILED,
e.getMessage()));
throw new PdfOcrTesseract4Exception(
PdfOcrTesseract4ExceptionMessageConstant
.TESSERACT_FAILED);
}
}
/**
* Gets list of text infos from hocr page.
*/
private static List getTextData(Element page,
Tesseract4OcrEngineProperties tesseract4OcrEngineProperties,
List txt,
Map unparsedBBoxes) {
final Rectangle pageBbox = parseBBox(page, null, unparsedBBoxes);
final List searchedClasses = Arrays.asList(OCR_LINE, OCR_CAPTION);
Elements objects = new Elements();
for (int i = 0; i < searchedClasses.size(); i++) {
Elements foundElements = page
.getElementsByClass(
searchedClasses.get(i));
for (int j = 0; j < foundElements.size(); j++) {
objects.add(foundElements.get(j));
}
}
return getTextData(objects,
tesseract4OcrEngineProperties,
txt,
pageBbox,
unparsedBBoxes);
}
/**
* Gets list of text infos from elements within hocr page.
*/
private static List getTextData(List pageObjects,
Tesseract4OcrEngineProperties tesseract4OcrEngineProperties,
List txt,
Rectangle pageBbox,
Map unparsedBBoxes) {
List textData = new ArrayList();
for (Element lineOrCaption : pageObjects) {
if (!lineOrCaption.text().isEmpty() && isElementConfident(lineOrCaption,
tesseract4OcrEngineProperties.getMinimalConfidenceLevel())) {
String hocrLineInTxt = findHocrLineInTxt(lineOrCaption, txt);
if (tesseract4OcrEngineProperties.getTextPositioning() == TextPositioning.BY_WORDS
|| tesseract4OcrEngineProperties.getTextPositioning() == TextPositioning.BY_WORDS_AND_LINES) {
for (TextInfo ti : getTextDataForWords(lineOrCaption,
hocrLineInTxt,
tesseract4OcrEngineProperties.getTextPositioning(),
pageBbox,
unparsedBBoxes)) {
textData.add(ti);
}
} else {
for (TextInfo ti : getTextDataForLines(lineOrCaption,
hocrLineInTxt,
pageBbox,
unparsedBBoxes)) {
textData.add(ti);
}
}
}
}
return textData;
}
/**
* Decides if lineOrCaption is confident or not given into account
* minimalConfidenceLevel property of {@link Tesseract4OcrEngineProperties}.
*/
private static boolean isElementConfident(Element lineOrCaption, int minimalConfidenceLevel) {
if (minimalConfidenceLevel == 0) {
return true;
} else {
int wconfTotal = 0;
int wconfCount = 0;
for (Node node : lineOrCaption.childNodes()) {
if (node instanceof Element) {
String title = ((Element)node).attr(TITLE);
Matcher matcher = WCONF_PATTERN.matcher(title);
if (matcher.matches()) {
String wconf = null;
try {
wconf = matcher.group(1);
} catch (Exception e) {
//No need to do anything here
}
if (wconf != null) {
wconf = wconf.replaceAll(X_WCONF, "").trim();
wconfTotal += Integer.parseInt(wconf);
wconfCount++;
}
}
}
}
if (wconfCount > 0) {
return wconfTotal / wconfCount >= minimalConfidenceLevel;
} else {
return true;
}
}
}
/**
* Gets list of words represented by text infos from hocr line.
*/
private static List getTextDataForWords(Element lineOrCaption,
String txtLine,
TextPositioning textPositioning,
Rectangle pageBbox,
Map unparsedBBoxes) {
List textData = new ArrayList();
if (txtLine == null) {
for (Element word : lineOrCaption.getElementsByClass(OCRX_WORD)) {
final Rectangle bboxRect = getAlignedBBox(word,
textPositioning, pageBbox,
unparsedBBoxes);
addToTextData(textData, word.text(), bboxRect);
}
} else {
List textInfos = new ArrayList<>();
final String txtLine1 = txtLine.replaceAll(NEW_LINE_PATTERN, "");
final String txtLine2 = txtLine1.replaceAll(SPACE_PATTERN, " ");
String[] lineItems = txtLine2.split(" ");
for (Element word : lineOrCaption.getElementsByClass(OCRX_WORD)) {
final Rectangle bboxRect = getAlignedBBox(word,
textPositioning, pageBbox,
unparsedBBoxes);
textInfos.add(new TextInfo(word.text(),
bboxRect));
if (lineItems[0].replaceAll(NEW_LINE_OR_SPACE_PATTERN, "")
.equals(getTextInfosText(textInfos).replaceAll(SPACE_PATTERN, ""))) {
lineItems = Arrays.copyOfRange(lineItems, 1, lineItems.length);
addToTextData(textData, mergeTextInfos(textInfos));
textInfos.clear();
}
}
}
return textData;
}
/**
* Gets list of lines represented by text infos from hocr line.
*/
private static List getTextDataForLines(Element lineOrCaption,
String txtLine,
Rectangle pageBbox,
Map unparsedBBoxes) {
List textData = new ArrayList();
final Rectangle bboxRect = getAlignedBBox(lineOrCaption,
TextPositioning.BY_LINES, pageBbox,
unparsedBBoxes);
if (txtLine == null) {
addToTextData(textData, lineOrCaption.text(), bboxRect);
} else {
addToTextData(textData, txtLine, bboxRect);
}
return textData;
}
/**
* Add text chunk represented by text and bbox to list of text infos.
*/
private static void addToTextData(List textData,
String text,
Rectangle bboxRect) {
final TextInfo textInfo = new TextInfo(text, bboxRect);
textData.add(textInfo);
}
/**
* Add text chunk represented by text info to list of text infos.
*/
private static void addToTextData(List textData,
TextInfo textInfo) {
String text = textInfo.getText();
Rectangle bboxRect = textInfo.getBboxRect();
addToTextData(textData, text, bboxRect);
}
/**
* Gets common text for list of text infos.
*/
private static String getTextInfosText(List textInfos) {
StringBuilder text = new StringBuilder();
for (TextInfo textInfo : textInfos) {
text.append(textInfo.getText());
}
return text.toString();
}
/**
* Merges text infos.
*
* @param textInfos source to merge
* @return merged text info
*/
private static TextInfo mergeTextInfos(List textInfos) {
TextInfo textInfo = new TextInfo(textInfos.get(0));
for (int i = 1; i < textInfos.size(); i++) {
textInfo.setText(textInfo.getText() + textInfos.get(i).getText());
Rectangle leftBBox = textInfo.getBboxRect();
Rectangle rightBBox = textInfos.get(i).getBboxRect();
textInfo.setBboxRect(new Rectangle(0, 0).setBbox(
leftBBox.getLeft(),
Math.min(leftBBox.getBottom(), rightBBox.getBottom()),
rightBBox.getRight(),
Math.max(leftBBox.getTop(), rightBBox.getTop())
));
}
return textInfo;
}
/**
* Attempts to find HOCR line text in provided TXT.
*
* @return text line if found, otherwise null
*/
private static String findHocrLineInTxt(Element line, List txt) {
if (txt == null) {
return null;
}
String hocrLineText = line.text().replaceAll(SPACE_PATTERN, "");
if (hocrLineText.isEmpty()) {
return null;
}
for (String txtLine : txt) {
if (txtLine.replaceAll(SPACE_PATTERN, "").equals(hocrLineText)) {
return txtLine;
}
}
return null;
}
}
© 2015 - 2024 Weber Informatics LLC | Privacy Policy