Re: Subtracting G channel from R and B may increase lossless compression?

OK, so I think a good idea would be use RGB decorrelation (R-G, B-G, G), decorrelation by YUV lossless conversion or none choosing the best for each line (or block), at the cost of a couple of bits more.

I'll try this solution as it seems to me a good one.

Thank you again Danilo

"cr88192" <cr88192@xxxxxxxxxxxxxxxxxx> wrote in message news:9feb1$465617ba$ca83a8d6$22326@xxxxxxxxxxxxx

"LiloLilo" <danilo.brambilla@xxxxxxxx> wrote in message news:4655e5ba$0$16029$5fc30a8@xxxxxxxxxxxxxxxxxx
Thank you very much. Very userfull link.

It seems that this kind of transformation would be quite different from the usual YUV. Do you think that this difference would be small? Is there any comparison between them?


actually, they are at least vaguely similar.


Suppose I have now converted my color space from RGB to YUV and let's speak again of channel differencing. The correct differencing now wold be U-Y, V-Y and Y?


no, this is a bad idea with YUV, because this particular relation is already implied by the transform itself (and as such, doing such an operation would only serve to re-correlate it...).

for example, with a typical YUV-based representation, if U and V are always 128, the image is greyscale.

or such...


Thank you all for help, you are great!


<snip>




.