quot;Compressibilityquot; check using SSIM like, better, really, should

andy

I have read the threads about the (relatively small) usefullness of compressibility checks with x264 (or any AVC, or xvid for what matters) but I was thinking that what one wants is a quot;qualityquot; check. I don't know much about SSIM but from the MSU test  video/code...arison_en.html it should be a much better predictor of the final quality of the compression. Obviously it's not gonna be really accurate, but it looks like it's much more accurate than PSNR and probably than the quot;-q 2quot; of xvid. And any kind of automatic suggestion, is better than none.

Let the flames start!  Actually, I'll give you some arguments for rebuttals:
1) Who cares about fitting movies on CDs or DVDs any more? We will move everything to HD pretty soon anyway.
2) SSIM and everything else that gives a number to the quality of a video file is useless, since it only counts what's best to the eyes of a single user, and everyone has different tastes.
3) Yeah, it's a good idea, why don't YOU do it!?

compressibility check is about COMPRESSIBILITY...
It's how much you can compress without loosing too much quality.
If you want to estabilish a quality meter then it's easier compressing directly with constant quantizer.
however you can do a compressibility check with h264 but you should estabilish some fixed parameters first (like number of reference frams, deblocking strenght and threshold, etc) that can influence compressibility.
The problem is: what are the appropriate values for those parameters?
you should manually evaluate how to set them and since they influence compressibility, if you set them wrongly the encoding quality will not be the quot;bestquot; and compressibility test will be useless...


Originally Posted by Sharktoothcompressibility check is about COMPRESSIBILITY...
It's how much you can compress without loosing too much quality.
If you want to estabilish a quality meter then it's easier compressing directly with constant quantizer.

It is easier, but constant quantizer (and correct mw if I'm wrong) uses too much bitrate on fast motion and dark scenes, and I guess SSIM should take those into account and be a more accurate rappresentation of the quot;qualityquot;.Originally Posted by Sharktoothhowever you can do a compressibility check with h264 but you should estabilish some fixed parameters first (like number of reference frams, deblocking strenght and threshold, etc) that can influence compressibility.
The problem is: what are the appropriate values for those parameters?
you should manually evaluate how to set them and since they influence compressibility, if you set them wrongly the encoding quality will not be the quot;bestquot; and compressibility test will be useless...

Well, usually I just use HQ-Slow , and in general a slower setting gives a better quality than a fast one. But that's something that I decide in advance, I use the compressibility just to decide at what resolution I should encode.