[poppler] Overprint implementation in postscript and splash device

[Albert Astals Cid]

El Diumenge, 12 de febrer de 2012, a les 08:43:58, Thomas Freitag va escriure:
> Hi!
> 
> As announced, here my overprint patch. It is an enhancement patch, a
> merge fix and a bug fix in one: an enhancement, because it now completes
> the implementation overprint mode and devicen in postscript, a merge
> fix, because it fixes some bugs in the overprint implementation in
> splash of xpdf 3.0.3 and has now the complete functionality (and more!)
> of my implementation back again and a bug fix, because it fixes the use
> of splash cmyk in postscript which never had worked.
> 
> 1. Overprint implementation in postscript
> To have a complete overprint implementation in the (pure) postscript
> device there were just two things missing: overprint mode and the
> implementation of the DeviceN colorspace in PostScript. I double checked
> my implementation with the Ghent Test Suite with GhostScript (device
> pdfwrite) and Acrobat X distiller, and all the tests now succeeds,
> either in Acrobat X distiller or in GhostScript. As overprint is a
> device dependent feature, it is up to the output device if it supports
> overprint and what features of overprint are supported, and often You
> have various configuration possibilities there. Nearly all PostScript
> output of the Ghent tests show now the desired results if converting it
> back to PDF with ghostscript pdfwrite, the implementation in ghostscript
> is complete. On the other hand a few tests failed when using Acrobat X
> distiller, all of them with the overprint mode switch. Funny, because
> overprint mode 1 is often also called the illustrator overprint mode
> because it was introduced by illustrator, but probably the destiller
> only handles EPS correctly which comes from illustrator :-)
> 
> 2. Overprint implementation in postscript if using splash rasterization
> Of course the postscript overprint implementation will only work if
> pdftops doesn't decide to use splash to rasterize it because of the use
> of transparencies in the PDF. But because overprint is device dependent
> I decided to spend an additional parameter "overprint" to pdftops
> (simular to pdftoppm). Switching it on (only available if compiled with
> SPLASH_CMYK, because overprint is only in CMYK colorspace showable) will
> use the overprint implementation in splash also if rasterizing the PDF.
> 
> 3. Overprint implementation in splash
> The overprint implementation in splash now uses the better designed
> interface of xpdf and therefore now also works in transparency groups.
> Thanks to the developper team of xpdf. I just fixed a small bug in it,
> where it defies the technical specification. Of course it is still in
> the nature of the splash implementation that it fails if CMYK values
> should overprint spot colors, because spot colors are converted
> immediately in their CMYK alternates. And in the implementation of
> overprinting spot colors over CMYK colors I made a small assumption to
> get it running which causes undesired results if this assumption fails,
> but I didn't want to implement more and more configuration switches. I
> still plan to implement a DeviceN output in splash and make a complete
> implementation there, but this is of course a bigger task (but doable).
> The overprint switch in pdftoppm is still only available if compiled
> with the SPLASH_CMYK directive.
> 
> If someone wants to know more about overprint, just contact me directly.
> Then I'll send the technical papers and/or hints to specifications.

Pushed to master.

Albert

> 
> Cheers,
> Thomas