Color & image issue

[mmettey]

Hi,

I am encoutering an issue when converting a specific PDF file.
I tried to generate this PDF with two different libraries. They look almost identical, but I have two different issues when converting them to SVG.

1st case - color alteration

When I convert this PDF :
page58-gs.pdf
I get this result :
page58-gs-1.svg

We can see that the left blue border is becoming red in SVG.

2nd case - removed images

When I convert the second PDF :
page58-mu.pdf
I get this result :
page58-mu-1.svg

Here, the color is right, but some images on the top right corner of the page disappeared.

I had these issues using the last CLI version of pdftosvg.net (1.5.0).

Thanks in advance for your help !

[dmester]

Hello,

Thanks for reporting this issue.

1st case

The blue border is using ICC colors. PdfToSvg.NET does not currently support ICC profiles and falls back to assuming they are RGB colors, which does not produce a good result here. The result is the same if opening the PDF in Chrome and Firefox.

Is it possible to change the color to RGB or CMYK as a workaround?

2nd case

The images are JPEG2000 encoded, which is currently not supported.

I was under the impression this format was practically dead. Did the library itself encode it to JPEG2000, or did you use a source JPEG2000 image?

[manfromarce]

It also happens to me that PDFs can contain JPEG 2000 or JBIG2 images if you have no control over the input file. Would it be possible to integrate libraries such as https://github.com/cinderblocks/CoreJ2K and https://github.com/nicholsab/JBig2Decoder.NETStandard to decode these formats?

[dmester]

Support for JBIG2 is included in the not yet released v1.6.0. I have put JPEG 2000 on the todo list for a future version.

[mmettey]

Thank you very much for clarifying the issue.

In the first case, the PDF was extracted from a full source PDF using Ghostscript. Since Ghostscript rewrites a new PDF instead of just splitting the pages, it appears to alter some elements, specifically the color types.

I confirm that in the second case, the PDF file was not altered by the library; it was simply split from the full PDF source using mutool (MuPDF).
I tried converting the same page using the full source PDF and the --pages parameter, but the result was the same. This suggests the source file was created using the JPEG2000 format for these images.

Given these findings, I'll explore options for adjusting the source file before conversion.
Maybe I can suggest to implement some kind of debug mode (with a new parameter --debug), where the CLI would return warnings when elements like JPEG2000 are found and not converted ?