[PDFBOX-4326] PDF with JPEG2000 image can't be rendered

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.0.11
• Fix Version/s: 2.0.13, 3.0.0 PDFBox
• Component/s: Rendering
• Labels: JPEG2000 JPXDecode JPXFilter
• External issue URL: https://github.com/mozilla/pdf.js/issues/10026
• Assignee: Tilman Hausherr
• Reporter: Tilman Hausherr
• Created: 29/Sep/18 11:43
• Updated: 02/Dec/18 17:06
• Resolved: 05/Oct/18 17:09

Description

There is an EOF exception:

    java.io.EOFException
    at javax.imageio.stream.ImageInputStreamImpl.readBits(ImageInputStreamImpl.java:687)
    at org.apache.pdfbox.pdmodel.graphics.image.SampledImageReader.fromAny(SampledImageReader.java:523)
    at org.apache.pdfbox.pdmodel.graphics.image.SampledImageReader.getRGBImage(SampledImageReader.java:217)
    at org.apache.pdfbox.pdmodel.graphics.image.PDImageXObject.getImage(PDImageXObject.java:418)
    at org.apache.pdfbox.pdmodel.graphics.image.PDImageXObject.getImage(PDImageXObject.java:400)
    at org.apache.pdfbox.rendering.PageDrawer.drawImage(PageDrawer.java:1021) 

The EOF exception is because SampledImageReader expects a stream that is 3x longer than actually appears, because the colorspace is said to have 3 colors instead of 1. This is because the JPEG2000 plugin returns an image with IndexColorModel which references three colors.

There are two ways to solve this and both worked: the slower is to convert to RGB which requires another change when reading the raster (because the raster is now made of integers), the faster is to force a gray colorspace. I use the later but keep the change I made in passing the integer RGB raster, we may need it some day.

I'm making the change only for the trunk for now, and to the 2.0 branch after release of 2.0.12.

Issue Links

• is duplicated by: PDFBOX-3246 Rendering PDF containing Jpeg2000 with red color
• relates to: PDFBOX-2204 Indexed color space in JPX