Hi,
I was looking for something that can reduce the colours in an image to the desired amount (e.g. 24 colours) so that the image looks as good as possible.
I think the ideal algorithm might place more of a weighting on colour ranges that take up more area of the picture… e.g. if there is an image with text and a gradient, it would give the gradient a lot of colours - to minimize colour banding (if dithering isn’t used) or speckling(?) (if error-diffusion is used).
So ideally it should be able to have error diffusion differing (or a superior dithering algorithm if it is available) as well as being able to have no dithering.
Ideally it should be free and also include source code but that isn’t a major priority.
It should also choose colours based on our visual limitations… e.g. it should have a higher weighting for brightness than it does for hue. And it could place a lower weighting on colours we can’t distinguish between very well - e.g. blackish colours.
Basically I am just concerned with optimizing the file-size… the speed of the program is completely irrelevant.
Thanks.
Frame differencing provides absolute minimum data rates; HVS (human visual system) fixed palette reduction algorithm provides excellent quality.
www.simtel.net/pub/pd/3543.html
www.digfrontiers.com/animDocs/Reduce.html
hyperarchive.lcs.mit.edu/HyperArchive/Abstracts/gst/grf/hvs-animator-lite.hqx.txt