2

I am trying to develop a lossless image compression algorithm. I know that YCbCr <-> RGB is practically lossy due to rounding errors, similarly Original Reversible Color Transform (ORCT) offer reversibility at the cost of storing extra bit for U and V component.

Since U and V are no way equivalent to Cb and Cr, the compression ratio differs greatly (I believe that this is due to their underlying cocktail of blending RGB in Cb and Cr).

Furthermore, I know that there exist techniques which require extra bits to accommodate reversibility (i.e. YCoCg-R etc). However I have tested YCoCg24, YUV (from ORCT) and GCbCr 1 but none of them come close to lossy YCbCr.

My question is that is there some reversible transform which approximate Cb and Cr since these two components play vital role in overall compression?

Before anyone blames me for not doing my homework, I should clarify that question is related to Lossless RGB to Y'CbCr transformation and JPEG: YCrCb <-> RGB conversion precision.

EDIT: To clarify that this is another question

My question is: does a transformation exist, that converts three eight-bit integers (representing red, green and blue components) into three other eight-bit integers (representing a colour space similar to Y'CbCr, where two components change only slightly with respect to position, or at least less than in an RGB colour space), and that can be inversed without loss of information?

Asif Ali
  • 81
  • 5
  • This does seem to be the exact same question/problem as https://stackoverflow.com/questions/10566668/lossless-rgb-to-ycbcr-transformation; it's not clear that you're going to get a better answer by asking it again :/ – Oliver Charlesworth Jan 27 '18 at 12:29
  • Dear Oliver, My question is to approximate CbCr components in reversible form while the provided link focusses solely on reversibility. (furthermore, I already mentioned that I implemented the transforms from in provided link). – Asif Ali Jan 27 '18 at 12:35
  • I understand that you've already tried it. What I'm saying is that that question asks for "a colour space similar to Y'CbCr ... that can be inversed without loss of information", which seems to be exactly what you're asking too. The accepted answer to that question is already pretty comprehensive. I get that it might be frustrating that there's apparently nothing better, but posting a duplicate question won't change that fact. – Oliver Charlesworth Jan 27 '18 at 12:38
  • Please see the edit: you will realize that my question is regarding the approximating CbCr (I am also willing to store additive bits too) while original question is about reversible transform in constrained bit-length. In-fact Cb and Cr components from GCbCr [1] are roughly equivalent of U and V from ORCT. – Asif Ali Jan 27 '18 at 12:49

0 Answers0