One-Clip-SSIM-Duel VP6.2 vs Xvid @ 1670 kbit

[kilg0r3]

Hi G&Gs

Just wanted to try out this SSIM Avisynth plugin method for image comparison. So I put Vp6 and Xvid against each other, and, it gave me some surpising results. i currently do not have the time to do more testing go into more details but Ithought it might be interesting for others. My surprise is due to the relation to Sagitaire's results, who is testing the lower midrange bitrates in his thread results

Clip
Sagitaire's Harry Potter 2 Trailer

Filesize
Aim: 25000KB
Xvid: 25002KB
VP6.2: 25812KB


The Settings

Xvid:
General: Qtype H.263, BVOPS:2/1.5/0, BVOP-Sensitivity 20; VHQ4, ChrMotion, Trellis, Defaults ...
Second Pass: I Frame Reduction 0 , Deafaults..

VP6.2
General: Bitrate: 1673kbit Mode: Two-pass Best, Local file VBR Sharpness 0 (6 gave worse results) Auto Keyframe
Advanced: Undershoot 90, Adjust Q: 2-330, Rate Varaiability 70 (bad?), Min Section 40, Max Section 400


The Results
Xvid: Average SSIM= 86.24
Vp6.2: Average SSIM= 81.88


The Encoding Script
LoadPLugin("C:\Programme\Avisynth\Plugins\MPEG2Dec3dg.dll")
Video=Mpeg2Source("D:\RIP\Test\hp2\source\1.d2v")
Video=Crop(Video,16,76,688,426)
Video=BicubicResize(Video,688,312)
Return(Video)


The Test Script
LoadPLugin("C:\Programme\Avisynth\Plugins\MPEG2Dec3dg.dll")
Video=Mpeg2Source("D:\RIP\Test\hp2\source\1.d2v")
Video=Crop(Video,16,76,688,426)
Video=BicubicResize(Video,688,312)

Clip2=AviSource("D:\RIP\Test\hp2\xvid.avi")
Clip3=AviSource("D:\RIP\Test\hp2\vp6.avi")
#return ssim(Video,clip2, "D:\RIP\Test\hp2\xvid.csv", "XVIDaverageSSIM.txt", lumimask=true)
return ssim(Video,clip3, "D:\RIP\Test\hp2\vp6.csv", "VP6averageSSIMvp.txt", lumimask=true)

The Script was loaded and 'played' in vdm 1.4.1
The Script was of course run twice, with the sharp sign in different poisitions
The script is not entirely correct. In the case of the Xvid avi I had to trim one or two frames off of VIDEO in order to get the right stream length. I don't really understand this because I am using the latest versions of Dvd2Avi and mpeg2dec. Of course I checked in VDM 1.4 that I had trimmed the correct frames.

Bye

[virus]

some random remarks:

- you entered the wrong size for VP6... if you want 1670 kbit/s, you need to set 1670/1.024=1630 Kbit/s in the dialog. That explains most of the oversize.
- what postprocessing has been used? VP6 with sharpness 6-10 at high bitrates works better with the deringer deactivated. As a general rule, if high values of sharpness don't perform well for a given clip/bitrate, don't assume that 0 is better. Sometimes 2-3 is the right choice.
- variability 70 should be fine. You may want to give a try at 60, but not lower.
- personally, I would have left Q at 4-56 for VP6. Don't know if this helps, though.
- for XviD, are you sure H.263 quant. is better than MPEG at this bitrate?
- please post some visual impressions too! :)

There's nothing wrong though. My experience shows that sometimes XviD is better than VP6, sometimes not. Both are excellent codecs, right now :cool:

[kilg0r3]

Originally posted by virus
some random remarks:

- you entered the wrong size for VP6... if you want 1670 kbit/s, you need to set 1670/1.024=1630 Kbit/s in the dialog. That explains most of the oversize. :stupid:
- what postprocessing has been used? since I played the clip in vdub in don't think any pp has been used. for vdub does not use the directshow filter. But I am not sure. IIRC in the direct show decoder config 'automatic' was ticked.
- for XviD, are you sure H.263 quant. is better than MPEG at this bitrate? this decision was based on a post by Sagitaire. Of course, I don't know. this post is partly thought as encouragement for others to do the same and compare custom matrices and the like.

! Guys, it is really easy to do if you read the introductory post for the plugin and know some of the avisynth basics. And, I am sure on your boxes it will run ahellufalot fastr than on mine. !
- please post some visual impressions too! :) i am still having exams. so, no sorry I don't have the time.

----------

A thousand, a thousand !! This is my thousandst (?) posting here.
thanks to everyone for being patient and tolerant enough to make this possible:)

[virus]

:stupid:
:confused: :sly:
can you explain that in plain English? I don't think I'm so stupid... the issue has been discussed early. And you've got a 3.2% oversize which is too much even for VP6.

[Soulhunter]

Originally posted by kilg0r3
! Guys, it is really easy to do if you read the introductory post for the plugin and know some of the avisynth basics. And, I am sure on your boxes it will run ahellufalot fastr than on mine. ! i am still having exams. so, no sorry I don't have the time.
THX, now I know how to do... ;)

Originally posted by kilg0r3
A thousand, a thousand !! This is my thousandst (?) posting here.
thanks to everyone for being patient and tolerant enough to make this possible:) Hehe, Im close behind you... :devil:


Bye

[SeeMoreDigital]

Originally posted by kilg0r3
...A thousand, a thousand !! This is my thousandst (?) posting here. What took you so long :D


Cheers

[Selur]

@virus: u missunderstood kilg0r3, I think that with :stupid:
he wanted to say that HE is the 'stupid' one, not you.
He was admitting his mistake ;)

Cu Selur

[kilg0r3]

Originally posted by SeeMoreDigital
What took you so long :D A wife, a child, exams and some issues specific to my autopoetic system :)

@ virus
Sorry for being ambiguous. Of course, with 'Im with stupid' I did _not_ mean that I am with you so that you were the one who were stupid. What I meant to say is that I am not with me but with stupid in the sense that stupidity is with me or better a part of me; probabely even the better part of me :)

Sorry for being not explicit enough earlier. :o

[SeeMoreDigital]

Originally posted by kilg0r3
A wife, a child, exams and some issues specific to my autopoetic system :) Well, I've had two of the first one and three of the second... if that's any sort of inspiration!

Anyways, back on track. Given what virus pointed out, are you going to generate another test?


Cheers

[kilg0r3]

nope. sorry. but just in case, what kinds of tests would you like to see?

[azerty@qwaerty]

My surprise is due to the relation to Sagitaire's results, who is testing the lower midrange bitrates in his thread results

1) 500-1000 Kbps for HPII trailer it's 30-60% compressibility interval for XviD and that corresponds to 99% of the encodings in 1cdr. In this interval VP6 or RV10 work very better than XviD and more extremist of XviD fan will say it to you

2) For VP6 in high bitrate (XviD ~quant2) choose this seeting:

General: Bitrate: 1630 kbit Mode: Two-pass Best, Local file VBR Sharpness 5 Auto Keyframe
Advanced: Undershoot 90, Adjust Q: 2-31, Rate Variability 100, Min Section 20, Max Section 800

3) No Post-Process actived for high bitrate: PP=0. Use this reg else PP doesn't work correctly in avisynth script ...


Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SOFTWARE\On2 Technologies\VFW Encoder/Decoder Settings]

[HKEY_LOCAL_MACHINE\SOFTWARE\On2 Technologies\VFW Encoder/Decoder Settings\VP60]
"STRCPUFREE"="0"
"strPostProcessingLevel"="0"

[HKEY_LOCAL_MACHINE\SOFTWARE\On2 Technologies\VFW Encoder/Decoder Settings\VP61]
"STRCPUFREE"="0"
"strPostprocessinglevel"="0"

[HKEY_LOCAL_MACHINE\SOFTWARE\On2 Technologies\VFW Encoder/Decoder Settings\VP62]
"STRCPUFREE"="0"
"strPostprocessinglevel"="0"


4) make comparative for high bitrate is not very interesting: even the MPEG2 with 1670 Kbps in 640*272 will obtain very good results with this trailer ... :devil:

[kilg0r3]

@ azerty@qwaerty

thanks for the advice :). I had the definite feeling that I have been missing a point here. The results were just too strange.

As I said i don't have the time to repeat the test. all the more because I'd have to reencode the files because i already deleted them. Stupid me ...4) make comparative for high bitrate is not very interesting: An opinion is an opinion is an opinion. Right? :)