Re: Fix (device interface) : Provide an access to image data filter for devices.

On Mon, Dec 29, 2008 at 8:35 AM, Ray Johnston <[email protected]> wrote:

> I think that for monochrome images, CCITT-G4 is MUCH faster than
> JBIG2, even when using the optional Luratech encoder. While
> JBIG2 compression can be superior to G4, I think we can get by
> with G4 and avoid the performance penalty and the requirement
> for the Luratech JBIG2 encoder.

I completely agree with Ray here. CCITT-G4 (or deflate) is the best
option for compressing bilevel images in the clist.

Further, as Ray suggests in a parallel thread, pass-through of already
compressed streams is an even better option. This shifts the
performance penalty for decoding expensive formats like JBIG2 and JPX
to the decode side, which will slow multithreaded rendering. However,
DCT and deflate-coded are much more common. In either case, avoiding
the extra decompress->recompress step and the reduced memory or disk
bandwidth should help.

 -r