Trent Piepho wrote: > > > I have both ways conversions: Y'CbCr <-> RGB, and both according to the newer > > colorspace FAQ coefficients/terminology are really Y'PbPr. I havn't tested the > > RGB -> Y'CbCr direction, but the Y'CbCr -> RGB converts bttv YUV data and the > > colors look normal. > > Are you getting the data directly from the driver, or from some application > that might be modifiying it? Directly from the driver - this is just a simple viewing test app. (in fact it's in my tarball on Gleicon's site): http://planeta.terra.com.br/informatica/gleicon/video4linux > Try capturing a colorbars test pattern in RGB mode and then YUV mode, and > convert using the Y'PbPr formula. Then try converting using the formula in > the bt848 datasheet. See which one looks more like the RGB output. You can > get the bt848 to generate a colorbars pattern internally, but you need to hack > the driver. That's good to know - I guess it would be easy to add a private bttv ioctl for it. I havn't done any side-by-side RGB vs YUV->RGB converted comparisons because previously I wasn't aware of the issue of alternate "Y'CbCr" definitions... I'm going to redo my conversions soon based on Justin's comment about integer multiply being cheaper than a lookup cache miss, and I'll do it then. Right now I've got more fun stuff to do! I've written a ground-up AVI for Linux library, and am at the point of implementing a few test CODECs :-) Ben
