1996.11.07 15:26 "Re: Tiff file informations", by Tom Lane

But this type of ordering is in contradiction with the way we store the YCbCr datas with YCbCrSubSampling Values of [2,2].

We have video images YCbCr in the format 4:2:2 (corresponding to ChromaSubSampleHoriz = 2 and ChromaSubSampleVert = 2) and the data are stored this way (the same that we read them) Y1, Cb1, Y2,Cr1, Y3, Cb2, Y4,Cr2 etc... in fact two values of luma for one couple of chroma (Cb,Cr).

Could someone ask this question: is the specification documentation false especially on page 93 or is the way we are storing datas wrong? It's very important for us to know, because we want to import TIFF files to our system.

4:2:2 is not [2,2] sampling, it is [2,1] sampling. You have chroma downsampled 2x relative to luma horizontally, but the same resolution vertically. A lot of people make this mistake, because the a:b:c notation is very unintuitive. (For reference, the other common sampling ratios are 4:1:1 == [4,1] and 4:2:0 == [2,2].)

If you are storing YCbCr data in that order in a TIFF file, then your TIFF file is unquestionably erroneous. Section 21 is perfectly clear that the TIFF sample order for [2,1] sampling is Y00,Y01,Cb,Cr.

If you are storing YCbCr data in that order in some format of your own devising, that's certainly a choice you can make. You will have to reorder the samples when converting to or from TIFF.

                        regards, tom lane