Codecs celerity ...

[Sagittaire]

Here small comparative of celerity enters the most recent versions of codecs their most powerful configurations. Same script avs was to use ...


RV9 04/06/2003
EHQ 80
2 pass: 1401 sec
5.88 fps

XviD Koepi
Ultra hight+VHQ4+bframe+chroma ME
2 pass: 1080 sec
7.65 fps

DivX Thahanea
Slow+bframe
3 pass: 2280 sec
5.42 fps

WMV9 VCM
Complexe+Quality max
2 pass: 1620 sec
5.09 fps

[Kurosu]

A PSNR (as the most classical and rather well accepted objective quality measure) result would be quite more meaningfull to me (I prefer quality of encoding over speed). Though I agree it's only easy (with avisynth) when source and output are both AVI/MPEG files (not RealMedia, therefore).

Please check documentation for Avisynth's Compare function if you are concerned by that point:
http://www.avisynth.org/index.php?page=Compare

Visual quality can be another point, but it's far too subjective (what if I prefer sharp images with noise while you prefer cleaned sources with too much blurring?) to satisfy me.

[Dark-Cracker]

keep in mind PNSR was NOT a quality indice.

[bilu]

Originally posted by Dark-Cracker
keep in mind PNSR was NOT a quality indice.

Compressibility and PSNR are both subjective quality indices.
Just for curiousity, do you know other quality indices?


Bilu

[Kurosu]

[edit: complete rewrote because that was just starting an useless fight]

PSNR can be considered as an objective quality measure (not based on perceptual matters). Check this definition for instance:
http://bmrc.berkeley.edu/courseware/cs294/fall97/assignment/psnr.html

It's right PSNR was criticized for being an artificial measure, not related to the perceived quality. In that regard, it may be interesting to read (for bilu, seeing his post) this pdf (ftp://ftp.crc.ca/crc/vqeg/phase1-docs/final_report_april00.pdf) hosted on a ftp (a word document is also available, but it's bigger and you have to own a non-free reader :)

My point was from the start (in order to avoid such controversy as brought by Dark-Cracker) to suggest PSNR as a measure, as the easiest to use, and a representative figure of the coding efficiency. Besides the fact Sagittaire's post doesn't seem to bring something valuable to the avisynth community (I may be totally wrong), I think that even the very basis of his post lacks some data: even for the same codec, a very similar filesize for 2 files doesn't mean they have the same quality (for instance reducing motion estimation precision and increasing quantizer).

Nota: I'm not at all criticizing how doom9 does his codecs comparisons, he does what he wants, it's up to you to decide whether those tests reply your questions about codecs' quality. And a visual quality is always more meaningful than a PSNR (although having both would be nice).

[edit]
@bilu
As I see now your post, I suggest you read the previous document. The subjective tests weren't completely accepted, and the conclusions seem to be still of little use, but you'll see some points where to start searching for other quality measures.

[bilu]

Your VQEG report has a weird address, can you check it please?

Thanks.


Bilu

[Kurosu]

Sorry for the inconvenience, I had to manually edit the link (as data was moved) and forgot to put back the host. Fixed by now. Happy reading :)

[bilu]

Just read pages 49 and 50, it said all I needed to know :)

To some extent, in our home-encoding hobby, also compressibility is a valid "safety-margin" (as I like to call it :) )

Thanks for the link.
Bilu