Need testing for Decomb 4.00 beta 1

[Guest]

Here is a first beta for Decomb 4.00:

http://shelob.mordor.net/dgraft/decomb400b1.zip

This version is changed as follows:

* User specified default parameters.
* Decimate() difference detection improved.
* Decimate metrics are normalized on the range 0-100% for easier tweaking.
* Decimate metrics can be displayed on the frame itself for easier tweaking.

Please note that the Decimate() parameters have changed radically and your old scripts may not behave the same way. Read the help file section on Decimate carefully! Also, for speed comparable to the previous version, set quality=0.

Feedback will be appreciated.

[stickboy]

From the Decimate help:
mode (0-1, default 2);)

I don't want to sound like a broken record, but have you discounted my "opposite-of-mode1" suggestion (http://forum.doom9.org/showthread.php?s=&postid=166675#post166675)? Were the results poor? Or is it still on the back-burner?

[Guest]

@stickboy

Sorry, I forgot to report the results of my experiment. The results were very poor. :(

But thank you for pointing out the error in my help file. :)

[Defiler]

I'll do a rip of Patlabor 2 using this right now and report back.

[Defiler]

Before I start, I just want to verify that this is a situation where I want "Telecide()" by itself, without "Decimate()"?
I encoded these (~400KB each) with DivX 5.02 Pro (which I don't normally use), because I know not everyone has XviD installed here.
If you'd like them in another format, please let me know.

I created this from a one-liner AVS file containing only:
mpeg2source("patlabor.d2v")
Source clip (http://hellninjacommando.com/misc/avisynth/pl2-divx5.avi)

Here's the same segment, with "Telecide()" added:
Test clip (Telecide) (http://hellninjacommando.com/misc/avisynth/pl2-telecide.avi)

..and finally, the segment with..
Telecide(guide=1)
Decimate(cycle=5,mode=2)
added:
Test clip (Telecide/Decimate) (http://hellninjacommando.com/misc/avisynth/pl2-both.avi)

[Guest]

I cannot properly analyse short source clips of divx encoded source. I need to see a VOB of at least 10MBytes with a lot of motion in it. Please post an URL where I can download it.

If it's a Telecide-only situation, it won't be very helpful because the Telecide functionality did not change, except for user specification of the default parameters.

[Defiler]

OK. I found something that I believe to be suitable.
VOB Test Sequence (13MB) (http://hellninjacommando.com/misc/avisynth/pl2-test.vob)
Sorry to make you examine another clip, but I wanted to make sure that this was a suitable test before wasting more of your time.

[Dali Lama]

Donald, I am not very good at testing out IVTC filters, because I don't really know what to look for; but to my untrained eyes, I see better decimation of a Bruce Lee movie: Chinese Connection using Decomb 4.0b. Areas that were choppy before are smoother. I may be wrong, but I also like the slower/higher quality setting in decimation.

The AVS:

Telecide(guide=1,post=false)
Decimate(cycle=5,quality=3)

Edit: Here is a link to a short clip of the scene in Huff format. (http://islam4you.hypermart.net/IVTC Sample.avi)

Good work,

Dali

[Guest]

@Defiler

Your clip should be processed simply by:

Telecide()

It appears to be a PAL->NTSC conversion done by blending rather than by 3:2 pulldown. There are two pieces of evidence that support this:

1. Presence of blended fields.
2. Lack of duplicates after Telecide().

You can see that there are no duplicates either visually or by adding Dup(show=true) after Telecide() and examining the metrics.

@Dali Lama

Thank you for the results. I would expect you to get better results with quality=3, because it does not subsample the frame. Previous versions of Decimate() always subsampled.

Note that eventually all my filters will support the show option to overlay tweaking information directly on the video. It is much more convenient that using DebugView (though it will still be supported as well to allow for logging).

[Guest]

Oh, forgot to say I just captured the season premiere of The Sopranos. I'll test the new version on it and report back.

EDIT:

Captured from DirecTV broadcast using Pinnacle DC10+ and AVI_IO. Processed with this script:

LoadPlugin("d:\avisynth\decomb.dll")
segmentedavisource("capture.avi")
Crop(0,64,0,-54)
Telecide(guide=1,gthresh=50)
Decimate(quality=3)
MSharpen(threshold=10,strength=100)

Encoded with DivX 5 Pro.

Result: superb!

[metallikop]

I'm still having the Decimate(cycle=5) problem I was having with 3.9 earlier. I'm currently running a decimate(mode=0) test to double check that it's the same error as before (but i'm 99% sure). Again, my encode crashes within the last 50 frames.

[Guest]

What is your full script? Does it happen with all clips? Does it happen when just playing the AVS in VirtualDub? What is the nature of the crash? etc. etc.

I just reviewed the other thread. You never answered my questions posed in an attempt to find the problem.

[Guest]

@metallikop

Playing a hunch, I recoded Decimate such that there is no chance it can ever request an out-of-range frame. Please tell me if the attached beta 2 helps with your problem.

[metallikop]

Thanks neuron2 I'll set up a pass right now for this.

[iago]

@neuron2

I'm considering to test the new beta 2 on a hybrid source (Baraka), for which the following script

LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
LoadPlugin("C:\PROGRA~1\GORDIA~1\decomb.dll")
LoadPlugin("C:\PROGRA~1\GORDIA~1\colorYUY2.dll")
mpeg2source("C:\BARAKA\BARAKA.d2v")
crop(4,4,704,472)
Telecide()
Decimate(mode=1,threshold=50)
BicubicResize(512,384,0,0.5)
ColorYUY2(Levels="TV->PC")

with the previous Decomb 3.91 didn't give me good results even going for 2CDs. Do you have any suggestions for improvement and for the test?

Thanks in advance and sorry if it sounds like a meaningless question due to my ignorance in IVTC.

kindest regards,
iago

[Guest]

with the previous Decomb 3.91 didn't give me good results even going for 2CDsI can only help if you tell me in what way your results were not good. As I have said before, I failed my Masters degree in mind reading. :)

When using mode=1, it's important to get the threshold parameter correct. Be aware it is defaulted to 0, and most applications will want to set a non-0 value. Use show=true to help tweak it.

[iago]

@neuron2

Well, actually I knew I would get such a reply and therefore I was preparing some screenshots to attach to better explain the situation ;).

OK, they are attached now. Thanks for your attention and time.

regards,
iago

[Guest]

@iago

Rather than approve the attachment, because it is very large, I'll just tell people it showed classic blending by the deinterlacer.

Please set blend=false as a parameter to Telecide() and see if you like that better.

And always give a text explanation when you post screenshots. Even there we can't always guess what you are pointing out. :)

[iago]

@neuron2

Thank you very much. I'll try it. I just wasn't sure a poor/inadequate explanation by me would define the situation clearly and therefore I preferred to send the screenshots only, believing that they could explain the issue better than me.

I'd already apologized for my ignorance in IVTC, which also involves describing what the real problem is :(. (I just didn't need to use IVTC before since all my sources were either PAL or only FILM >95% so far...)

Thanks again for your help and your time :).

best regards,
iago

[Guest]

@iago

Hey, no need to apologize for anything! Please let me know what your results are so we can try other things if needed.

[metallikop]

@neuron2

Your latest beta fixed my crashing problem. The .avs i was using was:


SetMemoryMax(40)
LoadPlugin("E:\PROGRA~1\GORDIA~1\mpeg2dec2.dll")
LoadPlugin("E:\PROGRA~1\GORDIA~1\decomb.dll")
MPEG2source("F:\DVD\movie\somemovie.d2v")
Decimate(cycle=5) # I also tried Decimate(mode=2)


I was experiencing the problem even after I took Telecide() out per your request, so I'm positive the latest beta fixed my issue.

Thanks again!

~k

[Guest]

@metallikop

Well, gosh, dude, isn't that just peachy keen! :)

You have entered the Graft Hall of Fame for contributing to the improvement of my software.

Something in your processing chain (MPEG2Source?) didn't like being asked to supply a frame beyond the end of the file. Oops. It was when you said that it happened always in the last few seconds that I had the hunch. Then I reviewed the GetFrame() calls and a few looked fishy. So I corrected them and also converted them all to a macro call. The macro checked for out-of-range calls and corrected them.

[iago]

Originally posted by neuron2
...Please let me know what your results are so we can try other things if needed.@neuron2

After your advice, I tried

Telecide(blend=false)
Decimate(mode=1,threshold=50)

which actually gave me a better result in terms of eliminating the blending effect (as I learnt BTW ;)) that occured in the previous encode as mentioned before.

However, later on I noticed that there were also other problems present in both encodes, both in the one with only Telecide() and in the one with (Telecide=blend=false), which I can describe as an unstable/stuttering picture and shaky, choppy playback in some still/non-motion scenes where only the camera moves very slowly. Only frame-analysis cannot detect the problem though; since it can be observed during playback.

Interestingly the best solution that eliminated both problems as well as increasing the quality of the encode came as selecting "Forced FILM" in DVD2AVI to create the d2v project file and work on the letterbox edition, with a resolution of 512*240, rather than the full-screen version; which possibly helped to allocate more bits to the movie, since AFAIK 70mm films are even harder to compress than the usual ones.

To repeat, the film was Baraka (70mm), NTSC1, about which DVD2AVI spat the following information: 4:3, 29.97fps, some parts NTSC 20-30% and some parts FILM <80, progressive(!). However in the previews both in GKnot and in VirtualDub interlacing/combing was clearly visible in motion parts...

Anyway, (after such a long description of the issue ;)) briefly, applying Forced FILM managed to solve my problems. I just wanted to inform you about it. And thanks again for your helpfulness and taking the time to read this post.

Kindest regards,
iago