On Thursday 10 January 2002 11:17 am, Alan Cox wrote: > Can you describe the format in more detail ? Well, I'll try... The luminance is extracted as the average value of the pixel being decoding and the one horizontally adjacent to it. The chrominance information is stored as lower bandwidth values stacked on top of these. The difference of adjacent pixels in odd rasters covers a line on the color wheel going from magenta/blue to green/yellow. Even rasters encode red and cyan. As a simple example, consider a 2x2 pixel matrix: 160 085 064 190 The pixel being decoded here would be represented by the value at 0,0. The luminance of the pixel is 122 ([160 + 85]/2). The magenta/blue - green/yellow chromaticity value is 75 (160-85), and the red - cyan chromaticity value is -126 (64-160). Pump this through an equation looking rather similar to a YUV-to-RGB conversion and you get a fairly high quality color image (well, after some gamma correction, rescaling and sharpening). In the next pixel over (where 85 would be the upper left) you have to flip the sign of the result of the chromaticity calculations of course. The project has an older sourceforge web page where the guy who reversed engineered the color encoding did a write up of his findings (http://homeconnectusb.sourceforge.net/indexold.html) and all our source code and other banter is up on that sourceforge project site.
