rate control by Zhihai's Paper

[Legend_Wolf]

Anyone have seen his paper"A linear source model and a unified rate control algorithm for DCT video coding". And in my opinion, his algorithm have to compute the whole frame coefficients( in h.263 like
a histogram), and it not fit for the rate control based on Macroblock.
"http://forum.doom9.org/showthread.php?s =&threadid=33204&highlight=zhihai" gave some information. However, I want to know how about the result.

[MfA]

For the moment only Michael has results, and he aint talking.

It is just for per frame quantizer selection, macroblock ratecontrol is a little more involved ... but mostly a seperate issue, as long as you dont consider dquants.

If you wanted to use it with dquants you could use the idea from thesis and assign the MBs to a couple of classes based on their SAD and assign a quant to each class, and determine dquants from that.

Personally I think the best use for single pass coding is fixed-distortion coding, and for two pass to calculate a global lambda for RD optimization.