Need some help from IVTC / fielding experts (like you neuron)

[Stigma]

Hi, and thank you for taking the time to help.

Iwe recently taken it upon myself to make some encodes of the R1 version of Hellsing, and iwe hit some snags i cant seem to figure out on my own. Hellsing isnt too clean when it comes to fielding, and thus i dont think a perfect IVTC is possible, so im not gonna ask about that. I got good enough results allready exept for my problem.

The problem occurs when there is panning of the camera directly up or down (which happens alot in anime). Neurons telecide decombing filter seems to have problems when this panning happens at a spesific speed (thus only a few ofthe panning straight up/down scenes are affected). Here are some screenshots hosted on my own server that will show you the problem. One good, as it should be, and one of the artifact. Note the jagged edges and "staircase" effect. it looks terrible in motion because the artifact also moves. The one good frame here is just one i found by cooincidence among all the bad ones.

ftp://help:help@stigma.serveftp.com/sample1(bad).jpg
ftp://help:help@stigma.serveftp.com/sample2(good).jpg
these pictures are taken with: telecide (chroma=true,blend=true,threshold=254) I could not seems to fix it by using any other paramters in the decomb help though.

also, here is a 50-frame and a smaller 10-frame (same scene) uncompressed huffyuv directly from the .vob (via DVD2AVI, no fielding operation), without any filtering of any kind. I assume this is as good as the vob itself, however i CAN fix the vob tomorrow if that will help solve the problem.

ftp://help:help@stigma.serveftp.com/error_sample.rar
ftp://help:help@stigma.serveftp.com/error_sample_small.rar
If more samples are needed from other scenes that create artifacts, then let me know. I just didnt have time for it right now ^_^

So far there HAS been a way for me to avoid this artifact though, by manually switching from telecide to IT.dll in the ares where telecide has problems (IT seems to mange these fine, but has other problems that prevent me from using it on all of the video). This is of course very timeconsuming, but better then nothing.

This sample however seems to throw even IT.dll off balance. Now im stuck with not even an emergency solution. So any way to fix this problem, even if its something like manually matching all the frames, then i would be grateful.

PS: I'l only be awake for another 20 mins or so, so most likely i cant respond until tomorrow morn unless theres someone who answers really quickly ^_^ Server will be operative for the clips&images though unless something unforseen happens.

-Stigma

[Guest]

Downloading now. I have that DVD and I think I know what your problem is.

Edit: Smilies are not allowed in thread titles. Please refer to the forum rules. Thank you.

And I am neuron2, but I do try harder.

[Guest]

The clip is too short and without enough movement to properly analyze it. Please provide a larger *VOB fragment* containing the scene leading up to this one as well as this one. Thank you.

EDIT: I picked the worst looking frame and then used Telecide's manual override feature to find the actual best match. Then I ran it through my "new generation Decomb" currently under development, which has a completely new field match engine, and it matched it correctly! The clip looks very good with the new design. You'll have to wait a week or two to get it, though. Follow it's progress by reading my journal at my website. This kind of clip is very difficult for a field matcher and has secured a place in my torture suite. Thank you for making it available. But please also get me the larger clip as described above.

[Stigma]

Allright, here it is. I took the time to cut the original vob this time in order to give you a much larger number of frames without making it such a huge download. This includes the problem scene, 2 previous scenes and a little snip after the problem scene. I figured i better be on the safe side ^_^

Heres the link:
ftp://help:help@stigma.serveftp.com/sample_vob_big.rar

I hope its enough to make any sense out of =)

A new decomb with a brand new field matching engine? Woohoo! Cant wait ^_^ But i hope you have some temporary solution for me to fix this problem though, even if it should involve a lot of work. Iwe never used the manual override before, so if thats the temporary solution, please try to give me as much helpful info as you can on how to use it in addition to that which is in your helpfile.

And thanks for taking the time to help

PS: Sorry about the smiley in the topic. I know iwe read that rule, but my fingers must have smiled before my brain reacted ^_^ It wont happen again.

[Stigma]

Oh by the way, just in case you happen to know since its within the same field of expertise:

Do you know how to turn off and on blending on IT.dll ? and how to set the threshold? when i use threshold=X and blend=true i dont get an error, so at first they seem like valid arguments, but no matter what i set them to there is no effect. Are they broken, or have i just overlooked something? (hope i dont have an old broken version or something)

[Guest]

Sorry, don't know about IT.

You can try this:

LoadPlugin("mpeg2dec3.dll")
LoadPlugin("decomb.dll")
loadplugin("uncomb.dll")
mpeg2source("big_sample.d2v")
uncomb()
fielddeinterlace(full=false)

That matches this scene better than Telecide, but it still gets one frame matched wrong. My new-generation decomb filter gets it all correct, but it is not released yet.

If it's just those few frames that bother you, use Telecide's override capability to match those frames manually.

[Stigma]

Uncomb? never heard of that one before, but thanks, il take a look at it or use decomb's manual override if i can figure it out.

Again thanks for your help. I will most certainly look forward to your new decomb.

[Stigma]

Hmmmm...

Well i gave uncomb.dll a try, but i can't get it to do any better then telecide, both when applying it directly to the trouble area or using it a bit before + trouble area. result is the same as telecide.

I also gave the manual override for telecide a try, but i get to about frame 4, and then i cant get a match wither either P C or N. Obviously im going at this the wrong way.

Could you decribe to me a step by step method to attack this problem frame-by-frame with manual override? Sorry to be such a pain, but im really stuck >_< and the decomb help file dosn't give too much detailed help on the subject.

EDIT: just some mis-spelling

[Guest]

The source simply doesn't allow perfect matches because the right fields aren't there. If you try all possible matches and don't find a perfect one, then it isn't there in the source!

You can prove it one way by using FORCE FILM and seeing the un-pulled-down frames. Or use the manual override process described below.

Uncomb() does do marginally better than Telecide() on this clip.

You're trying to recover something that is not in the source.

Manual override is simple. Remove Decimate. Then get the frame number of the frame you don't like. Then make a file like this:

58 p

Replace 58 with your frame number. Give this file as the 'ovr' parameter to Telecide(). Use p, c, and n to test all possible matches. You will see what I am talking about.

[Stigma]

so theres no way to recover it at all? ****! =(
So i guess even the new telecide wouldnt get it right then. Hmmz...

Well thanks for the help. Atleast now i wont spend days trying to do something that isnt possible...

*sob sob* what will happen now to my perfect Hellsing encode? ^_^

[Guest]

>so there's no way to recover it at all? ****! =(

You *can recover* the original source frames. You can't recover anything else.

>So i guess even the new telecide wouldn't get it right then. Hmmz...

Yes, it *does* get it right. It exactly recovers the source frames.