On Fri, 1 Sep 2006, chris clepper wrote:
On 9/1/06, chris clepper cgclepper@gmail.com wrote:
On 9/1/06, Mathieu Bouchard matju@artengine.ca wrote:
if you're in 8 bits per value, then "444" is 24 bpp, "422" is 16 bpp, and the others are 12 bpp, but those "bits per pixel" are assuming a "pixel" is defined only in terms of the Y channel. If you do it only in terms of UV channels, then everything is reversed, as "444" is 24 bpp, "422" is 32 bpp, and the others are 48 bpp.
It is important to state that all of the pixels are 24 bit in every YCbCr system. Each pixel can have a unique value even with the shared CbCr.
Well, I hope so!
If it didn't have a value of its own, it wouldn't be a pixel.
each pixel has a discrete 24 bits of information in YCbCr system. The subsampling means that pixels share some of the chroma (Cb and Cr) values and not that some pixels lack those color values.
Why is it important to make that distinction? There's no point in saying that pixels share chromas vs that pixels lack chromas. It depends on how you think of the pixel (e.g. is a pixel a point or is it a square?). To complicate matters, in practice, various types of interpolation are used, which leads the concepts of "sharing" and "lacking" to be a lot blurrier (pun intended).
BTW, congrats for the first mention of YCbCr in this thread. For anyone who doesn't know, Cb = blue chroma, Cr = red chroma, and those are synonymous with U and V, and also with B-Y and R-Y. Some people insist on distinguishing YCbCr, YUV and Y(B-Y)(R-Y) to mean slightly different things and may also use apostrophes and stuff, but in the current situation, one can't really expect a consistent notation from person to person.
_ _ __ ___ _____ ________ _____________ _____________________ ... | Mathieu Bouchard - tél:+1.514.383.3801 - http://artengine.ca/matju | Freelance Digital Arts Engineer, Montréal QC Canada