[TIKA-1729] OCR in PDF files

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.9, 1.10
• Component/s: config, parser
• Labels: java ocr parser pdf

Environment

Windows 7, 64-bit, JDK 1.8.0_51 64-bit
Windows 10, 64-bit, JDK 1.8.0_51 32-bit

• Assignee: Unassigned
• Reporter: Loris Bachert
• Created: 03/Sep/15 09:49
• Updated: 03/Sep/15 14:26
• Resolved: 03/Sep/15 14:26

Description

As described in this stackoverflow-post i'm having troubles extracting text out of scanned PDF files. By scanned PDF files i mean PDF files that consist only of images. Because each page is an image i can't extract them using a custom ParsingEmbeddedDocumentExtractor. I also tried using the setExtractInlineImages method of the PDFParserConfig but this didn't work aswell.
There was already a ticket regarding the OCR support and including the PDF file i'm using for my tests.
Here is a JUnit-test about my issue:

PDFOCRTest.java

@Test
public void testPDFOCRExtraction() throws IOException, SAXException, TikaException {
	File file = new File(filePath);
	InputStream stream = new FileInputStream(file);
	
	BodyContentHandler handler = new BodyContentHandler(Integer.MAX_VALUE);
	Metadata metadata = new Metadata();
	PDFParserConfig config = new PDFParserConfig();
	config.setExtractInlineImages(true);
	ParseContext context = new ParseContext();
	context.set(PDFParserConfig.class, config);
	
	PDFParser pdfParser = new PDFParser();
	pdfParser.setPDFParserConfig(config);
	pdfParser.parse(stream, handler, metadata, context);
	String text = handler.toString().trim();
	assertFalse(text.isEmpty());
}