Re: MMX speedup for Floyd Steinberg error diffusion

On May 10, 10:48 pm, Phil Carmody <thefatphil_demun...@xxxxxxxxxxx>
wrote:

You might be able to just do some kind of octree algorithm.
You could even hard-code the partitions.

(Octree's also possible in 256-colours too, of course. You can
even create a slightly optimised palate by using 2 passes and
crudely bin-counting on the 3 axes.)


I implemented 4 different quantizers, Octree, Median cut, Wu and
NeQuant ( the last 2 adapted from the FreeImage library ) and it seems
that Wu quantization gives the best results, followed by octrees, The
other 2 never quite get as good results as far as I could see.

What do you mean by "bin-counting on the 3 axes" ?

Regards
Vivek

.