DON'T USE Y'IQ OR Y'UV, USE Y'CBCR! Charles Poynton circa 1991 Don't use Y'IQ or Y'UV, use Y'CbCr instead: Y' = 0.299*B' + 0.587*G' + 0.114*B'; Cb = (0.5/(1-0.114))*(B'-Y'); Cr = (0.5/(1-0.299))*(R'-Y'); Y'CbCr is appropriate for systems that keep the chroma components separate. If you insist on multiplying (B'-Y') and (R'-Y') by sines and cosines and adding the products together as NTSC does to form a composite signal, then scale (B'-Y') and (R'-Y') by the textbook scale factors to form U and V, but this applies only if you are producing an NTSC signal to be conveyed on one piece of wire. If in forming the modulated pair you would like to reduce the bandwidth of one of the channels, then rotate the U and V pair 33 degrees and swap the axes, but again this applies only if you're going to eventually put the chroma components on one piece of wire. The most recent adoption of Y'IQ color space was 1953. Television has a long product cycle, and the technical documentation from the television technologists is long in being published (check the prop delay of SMPTE Journal articles, for instance). These are the reasons that YUV and YIQ keep cropping up even in the textbooks. Y'CbCr is the color space appropriate for modern component digital systems. Keep in mind that R', G' and B' are assumed to be gamma corected (i.e. tristimulus [linear light] values raised to the 0.5 power) prior to formation of Y', B'-Y', R'-Y'. C. -- Charles Poynton vox: +1 416 486 3271 fax: +1 416 486 3657 poynton@poynton.com [preferred, Mac Eudora, MIME, BinHqx] --