[Ffmpeg-devel] [PATCH] PNG 16-bit grayscale support

[Kostya]

On Wed, Oct 18, 2006 at 10:34:09AM +0200, Michael Niedermayer wrote:
> Hi
> 
> On Wed, Oct 18, 2006 at 08:07:29AM +0300, Kostya wrote:
> > On Sun, Oct 15, 2006 at 05:18:20PM +0200, Michael Niedermayer wrote:
> > > Hi
> > > 
> > [...]
> > > >  
> > > > +    PIX_FMT_GRAY16,     ///<        Y        , 16bpp big-endian
> > > 
> > > GRAY16 should be native endian, no realistic code besides memcpy and similar
> > > will be able to handle non native 16bit efficiently
> > > 
> > > if a PIX_FMT_GRAY16BE and PIX_FMT_GRAY16LE is usefull or not is something iam
> > > unsure about, id say we could always add them later if needed, but if you want
> > > to add them then iam fine with that too
> > 
> > A serach for documentation showed that formats that are known to support
> > 16-bit depth explicitly define it to be in "Network format order" (PNG spec)
> > or "MSB first" (PGM spec). But renaming format PIX_FMT_GRAY16BE seems logical.
> 
> ive no objections to adding PIX_FMT_GRAY16BE but theres one aspect you seem to
> forget and that is that this isnt only about raw or raw+generic compressor
> formats, normal codecs be it dct or wavelet based will always output native
> endian 16bit also filters will almost always use native endian 16bit as 
> its a matter or a+b vs. be2me(be2me(a)+be2me(b))

GRAY16 native endian needed for processing
GRAY16BE may needed for output (for PGM)
but as PNG explicitly defines GRAY16BE and may need to filter it, it will
take endian conversion in PNG decoding.

Maybe I'll add some more examples of 16-bit gray use (like PGM and JPEG LS)
It won't hurt FFmpeg to have full 16-bit (and 12-bit) grayscale support
 
> [...]
> -- 
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
> 
> In the past you could go to a library and read, borrow or copy any book
> Today you'd get arrested for mere telling someone where the library is
>