[jira] [Created] (TIKA-3002) Possible bug with OCR strategy AUTO

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

[jira] [Created] (TIKA-3002) Possible bug with OCR strategy AUTO

David Eric Pugh (Jira)
Patrick Herber created TIKA-3002:

             Summary: Possible bug with OCR strategy AUTO
                 Key: TIKA-3002
                 URL: https://issues.apache.org/jira/browse/TIKA-3002
             Project: Tika
          Issue Type: Bug
          Components: ocr, parser
    Affects Versions: 1.22
            Reporter: Patrick Herber

For performance reasons, I would like to activate the OCR scanning only when necessary. I therefore tried to set the OCR strategy to "AUTO".

However I see that also for "normal" PDF files (where no OCR should be required), OCR is performed and this not also slows down the application but (more important) results in doubling the resulting text.

Trying to understand how this works, I think I may have found a possible error in the class *AbstractPDF2XHTML*. There, in case of selected OCR Strategy AUTO, on line 404 is checked the total number of characters found on the page, if this is lower than 10 OCR is performed:

} else if (config.getOcrStrategy().equals(PDFParserConfig.OCR_STRATEGY.AUTO)) {
    //TODO add more sophistication
    if (totalCharsPerPage < 10 || unmappedUnicodeCharsPerPage > 10) {

The logic is correct, but unfortunately at the beginning of the method (line 361 and 362) the two variables checked on this line are reset to 0, so this conditions will always be true.

I would suggest to move the reset of the two variables inside a finally block at the end of the method.

This message was sent by Atlassian Jira