[
https://issues.apache.org/jira/browse/PDFBOX-3622?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15735751#comment-15735751
]
Tilman Hausherr commented on PDFBOX-3622:
-----------------------------------------
It's difficult to understand what happens. The input values are between 0 and
100 (ok), and the results returned by eval() are in the typical range of LAB
colors. So it's wrong to multiply them with 255. But even passing such values
directly doesn't work, maybe the raster doesn't accept "exotic" ranges. I
suspect I should do the reverse of what is done in PDLab.toRGBImage() (i.e.
"encoding" into a 0..255 range) but this would make it slower and more complex.
I've implemented an alternative converter which could be used for all
colorspaces, but is now used only for Lab. For other colorspaces the existing
path is taken, which may or may not be faster.
This change also brings a correct rendering of PDFJS-7138.pdf (
https://github.com/mozilla/pdf.js/issues/7138 ). Their problem is probably
different than ours, as they don't have trouble with the main PDF of our issue.
> DCTDecode image with separation colorspace not rendered correctly
> -----------------------------------------------------------------
>
> Key: PDFBOX-3622
> URL: https://issues.apache.org/jira/browse/PDFBOX-3622
> Project: PDFBox
> Issue Type: Bug
> Components: Rendering
> Affects Versions: 2.0.3
> Reporter: Tilman Hausherr
> Labels: LabColorspace, Separation
> Attachments: PDFBOX-3622-reduced.pdf, PDFBOX-3622.pdf
>
>
> PDF reported by Ford Wang in the mailing list. The "herbs" have weird purple
> dots. The cause are DCTDecode (jpeg) images with a separation colorspace,
> e.g. {{Root/Pages/Kids/\[0]/Resources/XObject/Im5267}} . The image itself is
> a grayscale image, i.e. it has only one color component. The separation
> colorspace is a green (PANTONE 355 C), which is based on an Lab colorspace.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
