Alexander Lubyagin
2011-Apr-04 09:39 UTC
[theora-dev] RLE for DC/AC-coefficients separately for the blocks or at whole?
It is proper way - to code DC/AC-coefficients for all macro-blocks at whole? I think that coding for DC/AC-coefficients for blocks separately gives lesser RLE-compression ratio. It is right? For example, at first we code DC-coefficients for all blocks. Compress they by RLE. Then, AC(1) coefficients. Compress it by RLE. And ... AC(63) coefficients for all blocks. Compress by RLE. It is better for RLE-coding?
Timothy B. Terriberry
2011-Apr-04 09:45 UTC
[theora-dev] RLE for DC/AC-coefficients separately for the blocks or at whole?
> I think that coding for DC/AC-coefficients for blocks separately gives lesser RLE-compression ratio. > It is right?Whether or not it's "right", the format does not give you the option. The token system defined in the specification has two ways of exploiting runs: runs of zeros _within_ a single block, and runs of EOB tokens _across_ multiple blocks. You can't code runs of zeros which span blocks, except through EOB tokens.