View Full Version : About matrix
Gibuk
10th January 2006, 10:50
excuse me: I have saved project d2v of both orignal and encoded vob (by DGDecode), started Seethedifference script (by CCE) but it closed and I get a compare.wav file to 0Kb.
Where Am I wrong?
manono
10th January 2006, 13:48
I'm running some tests now. I had made another mistake. The original .avs was for AVI, and AviSource, and I had changed it to MPEG2Source and D2V. But I forgot to add the DGDecode.dll into the script. Here's what mine says, and you can modify it for your paths:
LoadPlugin("F:\DivX Stuff\GKnot\DGDecode.dll")
V1 = MPEG2Source("K:\LOTR-EE-ROTK-Chap14\Tests\Original.d2v")
V2 = MPEG2Source("K:\LOTR-EE-ROTK-Chap14\Tests\Fox.d2v")
Sub = V1.Subtract(v2)
Substrong = Sub.Levels(122,1,132,0,255)
Return StackVertical(StackHorizontal(v1.Subtitle("original"),v2.Subtitle("encoded")),StackHorizontal(Sub.Subtitle("Difference"),Substrong.Subtitle("Difference Amplified")))
Since my tests use the same number of frames, I deleted the "FrameAdjust" lines. Since I'm not using any audio either, I deleted the "False" in the original script also. If you use this script with adjustments for your paths and D2V names, it'll work now. Sorry for my carelessness. I'll try and post some pics later, if I can figure out how ImageShack works.
That's a pretty cool script that guy came up with.
manono
10th January 2006, 17:30
http://img215.imageshack.us/img215/4415/originalframe8aj.th.jpg (http://img215.imageshack.us/my.php?image=originalframe8aj.jpg)
http://img215.imageshack.us/img215/1016/originalfox8dv.th.jpg (http://img215.imageshack.us/my.php?image=originalfox8dv.jpg)
http://img215.imageshack.us/img215/774/originalstandard6qy.th.jpg (http://img215.imageshack.us/my.php?image=originalstandard6qy.jpg)
http://img215.imageshack.us/img215/8058/foxstandard2zd.th.jpg (http://img215.imageshack.us/my.php?image=foxstandard2zd.jpg)
The first pic is the frame from which the comparisons were taken. That particular frame is a P-Frame using this matrix:
08 08 10 11 13 14 15 17
08 08 11 12 14 15 17 19
10 11 13 14 15 17 17 19
11 11 13 14 15 17 19 20
11 13 14 15 16 18 20 24
13 14 15 16 18 20 24 29
13 14 15 17 19 23 28 35
14 15 18 19 23 28 35 42
08 09 09 10 10 11 11 12
09 09 10 10 11 11 12 12
09 10 10 11 11 12 12 13
10 10 11 11 12 12 13 14
10 11 11 12 13 13 14 14
11 11 12 12 13 14 14 15
11 12 12 13 14 14 15 16
12 12 13 14 14 15 16 17
The R1 DVD of LOTR-ROTK-EE uses CCE's Standard matrix with AQ turned on, and that particular GOP used the Standard matrix with the numbers halved. The ave. Q of the entire original chapter was 5.62. When reencoding to 62.6% of the original size of the original M2V at a bitrate of 3000, the Fox matrix had an ave Q of 7.79, and the Standard matrix had an ave. Q of 3.67. I used the script from the previous post, substituting the different D2Vs.
The second pic is a comparison of the original and the same frame using the Fox matrix. The third pic is a comparison of the original with the Standard matrix, and the fourth pic is the Fox matrix compared with the Standard matrix.
One thing that jumped out at me was the difference between the original and even the Fox matrix pic. Although we all know it already, just the act of reencoding, no matter how good the settings, causes degradation, and reencoding for 62% of the original size (typical, I think when reencoding long movies, or normal movies keeping some extras), of course, causes further quality loss. I don't want to influence anyone in their conclusions, if any, so I'll suggest that anyone interested, grab the pics, put them into a folder, and run the Windows Slideshow, or any other way you know to see them one after the other. I've labeled each of the pics, so it'll be easy to tell them apart.
I may try again with a different film. I wasn't quite happy with the fact that the original scene used a lower quality matrix than Fox, and that I had chosen a low-action scene on which to test. I'll be happy to see any results anyone else might come up with.
archaeo
12th January 2006, 03:00
manono,
Finally, after many trials I had success in running the 'seethedifference' avs script :D
Found out my errors were from running an earlier version (v1.5) of VDub, since YV decompression wasn't supported until ver 1.6. (of course I got that valuable tidbit from a doom9 search) Anyway, it is an excellent way to do side by side pre-comparisons on matrix encodes, and I'll be using it to help make future decisions on matrix choices. The 'difference amplified' feature is also real nice.
-thanks for the help :)
Rockas
12th January 2006, 14:11
Ok... now I'm feeling lost :D
What is the purpose of 'seethedifference' (maybe this can be usefull for the future RME release - without the 'invalid Path' bug... I hope :))?
I know you probably mentioned on previous posts but my english plays tricks on me and I must have missed it :(
archaeo
12th January 2006, 14:36
Ok... now I'm feeling lost :D
What is the purpose of 'seethedifference' (maybe this can be usefull for the future RME release - without the 'invalid Path' bug... I hope :))?
I know you probably mentioned on previous posts but my english plays tricks on me and I must have missed it :(
Hello Rockas, long time :)
The purpose is that in this script, one can see the results of two (or more) of their encodes in a side-by-side video stream that makes it very easy to distinguish differences, say for example, between two matrix choices, two bitrate levels, or two filter choices... Or between the original stream and one of these. One of the best things is that I don't have to run one, wait, then run the next. That makes it harder to see the more subtle differences. By comparing two MPEGS in a single video stream, it is very easy to get an instant quality comparison.
Rockas
12th January 2006, 21:03
Well... I think I'll have to make some test... that's a great addon...
but now I have a doubt (sorry to ask this, but my test video computer is in the mechanic :) ) ... how does it display the two samples... with full resolution? how does it make it fit on the screen? resizes?
archaeo
12th January 2006, 22:47
Well... I think I'll have to make some test... that's a great addon...
but now I have a doubt (sorry to ask this, but my test video computer is in the mechanic :) ) ... how does it display the two samples... with full resolution? how does it make it fit on the screen? resizes?
It is full resolution. When viewing all four displays in 'seethedifference' (original, encoded, difference, difference amplified) it did not fit completely on my 17" monitor. There's definitely a need to either crop or resize for those with smaller displays. I am currently trying some crop and lancos resize commands to see if I can get it to a good size for my monitor.
Earlier today, I simplified the script to just show two streams, which are labelled 'original' and 'encoded'. It works very nicely to just show a basic side by side, for a an easy way to just juxtapose the two. In Vdub, they show at a good size to easily see the differences, without needing to do much resizing. Here it is:
#sidebyside
LoadPlugin("C:\Program Files\DVDTools\DGDecode\DGDecode.dll")
V1=MPEG2Source("F:\testclipRB\D2VAVS\V01.D2V")
V2=MPEG2Source("F:\testclip_filtr\D2VAVS\V01.D2V")
StackHorizontal(v1.Subtitle("original"),v2.Subtitle("encoded"))
Rockas
12th January 2006, 23:03
I don't think you should use resize... just a thought... wouldn't it be better with a crop?
then you'll have the original resolution to compare, right? ;) - just a thought... as I said I can't test it at the moment.
manono
13th January 2006, 01:00
I agree that a resize in the script might not be best. However, depending on your version of VDub(Mod) you can right-click the screen and resize by 4:3 or 16:9. If 16:9, you might want to StackVertical. You can also play the script in a lot of players, and some can also resize to 4:3 or 16:9. Media Player Classic can both play the script and resize it by 4:3 or 16:9. It can also pause and advance a frame at a time for closer examination. I'm sure other players can do the same thing.
archaeo
13th January 2006, 01:08
I don't think you should use resize... just a thought... wouldn't it be better with a crop?
then you'll have the original resolution to compare, right? ;) - just a thought... as I said I can't test it at the moment.
Yes, agreed. Cropping does clip some of the image away, but it doesn't reduce the visibility of artifacts, which was the point of this. Actually Rockas, I will also have to wait a week till I get back to this, since I'm taking a little vacation to the caribbean, starting tomorrow :cool:
I'll be glad to test more w/RME in mind when I return. :)
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.