On Wed, 2007-10-03 at 14:10 +0200, Herman Robak wrote: > On Wed, 03 Oct 2007 05:26:54 +0200, Graham Evans > <[EMAIL PROTECTED]> wrote: > > If the codec returns uncompressed 4:2:0 or 4:1:1 data, then the temporary > shouldn't just repeat the colour pixels in a 2x2 or 4x1 pattern. That's > not upsampling!
I was originally repeating chroma pixels myself, then after hearing this kind of talk on #cinelerra, I tried upsampling the chroma with linear interpolation. Result? Lots of small border artefacts of strange colours. To those who believe pixel duplication is not the way to go, can you point me to some digestible and specific algorithms (and preferably some straightforward reference code as well) that gives better upsampling? Better yet, something not polluted by patent encumbrance. My google searches for 'chroma upsampling' point me mainly to patents websites. Cheers David > It is not > visible > to the naked eye, since our eyes are rather poor at resolving small specs > of > colour. This is why chroma subsampling is acceptable in the first place. > But once a mask is derived from this coarse chroma signal, it gets ugly! > > That said, for a chroma keyer to be usable outside well-lit studios with > smooth greenscreens, it needs a little more features than chromakey-hsv. > Graham has suggested a few. > _______________________________________________ Cinelerra mailing list Cinelerra@skolelinux.no https://init.linpro.no/mailman/skolelinux.no/listinfo/cinelerra
