Log in

View Full Version : mpeg 2 encoder roundup

fjhdavid
4th July 2008, 17:14
I know the subject has been already discussed and that PSNR and SSIM are just mathematics approach, but did you see this mpeg 2 encoder roundup?

http://www.pate15.de/mpeg2-mpeg4/metriques.html


it looks like HC is better than CCE...even when you are looking to the pictures provided...

any comments?
jdobbs
5th July 2008, 01:02
Why would they be comparing XVID to MPEG-2? It's apples and oranges. I'll be honest... when I see a report that says QUENC and TMPGENC is better than CCE I have serious doubts as to any of its validity. Not that I don't like them (especially QUENC) but my experience in terms of quality is very much to the contrary.
Sharc
5th July 2008, 08:42
Why would they be comparing XVID to MPEG-2? It's apples and oranges.
The comparison is done at equal bitrates (file sizes), so I think it should be valid from this point of view(?).
For my eyes XviD lost details/smoothed the picture even at high bitrates, probably depending on the matrix used (H.263?).
Comparison on single (B-) picture basis is of limited value for the overall visual impression however. Ocoder (Procoder?) changed the color for example (colormatrix?)
Sharc
5th July 2008, 20:23
For encoder comparison I personally prefer to do it by means of displaying 2 clips side by side. Example script for comparing an Xvid encode with a HC encode:

#StackHorizontal(clip1, clip2)
V1=ffmpegsource("J:\Movies\310 To Yuma\My movie_.divx").Crop(0, 0, 360,416)
V2=ffmpegsource("D:\Movies\DVD Rebuilder\DVDFILES\VIDEO_TS\My movie_.VOB").trim(1462,0).Crop(0,80, 360, -80)
StackHorizontal(v1.Subtitle("XviD"),v2.Subtitle("HCenc"))

The trim command in V2 is used to synchronize the 2 clips.
jdobbs
5th July 2008, 20:38
That's good as long as you don't stop and compare frames. The problem with doing it that way is that you may be comparing an I-Frame to a B-Frame if you look at it frame by frame. The only way to compare a motion video is to play it back with motion.
45tripp
6th July 2008, 04:41
I'll be honest... when I see a report that says QUENC and TMPGENC is better than CCE I have serious doubts as to any of its validity. Not that I don't like them (especially QUENC) but my experience in terms of quality is very much to the contrary.

care to elaborate?
Sharc
6th July 2008, 07:33
That's good as long as you don't stop and compare frames. The problem with doing it that way is that you may be comparing an I-Frame to a B-Frame if you look at it frame by frame. The only way to compare a motion video is to play it back with motion.
Yes, for frames analysis one should at least step through a full GOP before making a quality judgement. Btw. is there a script command or similar to display the type of frame (I,B,P)? Info() seems not to deliver this information.
45tripp
13th July 2008, 16:18
Yes, for frames analysis one should at least step through a full GOP before making a quality judgement. Btw. is there a script command or similar to display the type of frame (I,B,P)? Info() seems not to deliver this information.

to avisynth a frame is a frame.
i've not seen a plugin for this. if you run into it, or make it, let me know.

you'd need an external read on the file to create a stats file and then a plugin to import into avisynth.

or you can just manually look at gop sequences with an external program and refer to it when needed. click click.

Why would they be comparing XVID to MPEG-2? It's apples and oranges. I'll be honest... when I see a report that says QUENC and TMPGENC is better than CCE I have serious doubts as to any of its validity. Not that I don't like them (especially QUENC) but my experience in terms of quality is very much to the contrary.

i'm still interested in the particulars of your experience.


tripp