Edited wiki page through web user interface.

import/raw
castano 17 years ago
parent 574d749720
commit e5661ae45c

@ -56,4 +56,4 @@ Note that signed components are packed as unsigned integers, so they need to be
A simple way of obtaining higher quality compression is also using the YCoCg color space, but encoding the Y component in the in the alpha channel of a DXT5 block, and the !CoCg components in the RGB channels. That was originally suggested by Waveren in [http://www.intel.com/cd/ids/developer/asmo-na/eng/dc/index.htm Real-Time DXT Compression]. This approach was described in more detail by Waveren and Castano in [http://developer.nvidia.com/object/real-time-ycocg-dxt-compression.html Real-Time YCoCg-DXT Compression]
[http://www.humus.ca Emil Person (aka Humus)] also has [http://www.humus.ca/index.php?page=3D&ID=68 a demo] that implements the same idea, but using YCbCr color space instead.
[http://www.humus.ca Emil Person (aka Humus)] also has [http://www.humus.ca/index.php?page=3D&ID=68 a demo] that implements a similar idea, but using YCbCr color space instead. He is actually using two textures, either one ATI1N (DXT5A) texture to store the luma (Y) and one ATI2N (3Dc) texture to store the chroma (CbCr), or one DXT1 texture to store the luma and one DXT5 (really DXT5nm) texture to store the chroma. The second texture that stores the chroma is a quarter the size to achieve a compression to 6 bits per pixel, as opposed to 8 bits per pixel for YCoCg DXT5. Although the compression ratio is good, you end up with two texture lookups.

Loading…
Cancel
Save