[Ffmpeg-devel] Bug in snow revision 1.50 - broken lossless mode

[D Richard Felker III]

On Wed, Apr 13, 2005 at 10:00:00PM +0200, Michael Niedermayer wrote:
> Hi
> 
> On Wednesday 13 April 2005 21:26, D Richard Felker III wrote:
> > On Wed, Apr 13, 2005 at 12:13:53PM +0200, Michael Niedermayer wrote:
> > > Hi
> > >
> > > On Wednesday 13 April 2005 06:03, Yartrebo wrote:
> > > > Michael, those rescaled IDWT coefficients (revision 1.50) appear to
> > > > have broken lossless mode quite bad. It's all fuzzy now and is similar
> > > > in quality to quant=10 or 20. It happens for all inputs I've tried.
> > >
> > > the regression tests worked yesterday and they do test lossless with the
> > > 5/3 wavelet
> > > the 9/7 wavelet is with the scaling merged not bijective with fixed point
> > > numbers, its bijective if rational or real numbers are used, but as the
> > > 5/3 wavelet is superior for lossless (at least it was in my tests) thats
> > > really irrelevant
> > > but then again, dont hesitate to look at the lifting steps for the 9/7
> > > wavelet, maybe you can come up with a better solution so that this
> > > dissadvantage can be avoided
> >
> > is the problem just that we have too few bits? because fracbits count
> > certainly be increased... i.e. we never seem to use more than 24 bits
> > or so out of the 32...
> 
> seems you missunderstand how lossless wavelet compression works
> the trick is to use 0 fractional bits, and no quantization or dequantization, 
> now the new 9/7 wavelet does a rescaling step by 4/5 during encoding and 5/4 
> during decoding which obviously wont be lossless that way, OTOH you can 

couldn't you scale by 8/5 and 5/8 instead so that the mapping would be
injective?

rich