[PDFBOX-4484] Some JBIG2 images are corrupted when subsampling is enabled

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.0.9, 2.0.10, 2.0.11, 2.0.12, 2.0.14
• Fix Version/s: 2.0.15
• Component/s: Rendering
• Labels: sideeffect
• Assignee: Tilman Hausherr
• Reporter: Pete Nattress
• Created: 12/Mar/19 15:42
• Updated: 12/Apr/19 14:39
• Resolved: 13/Mar/19 17:37

Description

We have observed problems when rendering some PDFs containing JBIG2 images if the subsampling option is enabled.

It is possible to reproduce this with the file I have attached using the CLI:

$ java -jar pdfbox-app-2.0.14.jar PDFToImage -subsampling test-jbig2.pdf
Mar 12, 2019 3:17:09 PM org.apache.pdfbox.pdmodel.graphics.image.SampledImageReader from1Bit
WARNING: premature EOF, image will be incomplete

In the output, the extracted image is distorted and has incorrect dimensions. This does not appear to affect all JBIG2 images, for instance the file "ItDoesntWorkScan.pdf" attached to PDFBOX-1067 contains a JBIG2 image and is rendered correctly with the above code. PDFs containing images encoded with algorithms other than JBIG2 also appear to be OK, although I haven't tested this exhaustively.

We have worked around this problem for now by disabling the subsampling option.