I been struggling for a while now, tring to fix this problem with my dvds that I'm trying to encode to put in my Hard drive. Exhausted, I'm turning to get expert help.
To begin with, I want to make the video progressive (even if I lose quality) but I'm not even sure that I'm using the proper way to deinterlaced/IVTC this video. The video goes 4 progressives and 1 interlaced (4:1 pattern), is this telecine or interlaced?
Look at the samples below.
A simple IVTC seems to get rid of the interlaced frame ( TFM().tdecimate() ). The problem is that I find here and there jaggy lines, that looks like aliasing artifact or deinterlacing artifact, but those were not introduce by the IVTC and were present before... maybe when they processed it to dvd ?
Anyway I am trying to find a way to stabilize those line, I found that using QTGMC( Preset="Slow", InputType=1 ) after the IVTC give a great result, not perfect but good enough. Which bring me to my second problem. From the moment I use QTGMC with this video, I get those weird artifact that looks like '' shadow lines'' or ''motion lines''blended into the frames, but you have to go frame by frame to really notice them.
Even if I skip the IVTC, and just use QTGMC (), TempGaussMC() or even animeivtc (mode=3). I notice those shadow/motion lines apearing all over the video. Which really bothers me since QTGMC does a great job of stabilizing the video.
So my questions is :
Is IVTC the way to go with this video ? QTGMC ().srestore() does the trick too, but comes with those F...artifacts.
Is there an other way to stabilize those jaggy lines ? (you'll find and exemple of those jaggy lines at the beginning and close to the end of the first sample)
And what's wrong with QTGMC, is there a way to configure it to not do those blended shadow/motion lines artifacts ? (the second sample is a good exemple when applied with the filter)
The sample where not process/encoded so you'll have to apply TFM().tdecimate() + QTGMC( Preset="Slow", InputType=1 ) to see what i'm taking about.
I would really appreciated your help ! Thanks
+ Reply to Thread
Results 1 to 15 of 15
I just realized that the sample I provided were in 23.976 fps. Somehow when I produce the sample with Dgindex it comes in 23.976fps, while the d2v I work with is at 29.970. Nontheless, the problem with QTGMC and the jaggy lines still remain... and I just don't know how to provide the file has I work with at 29.970 ??
Best link I could find with a quick search that mentions this technique.
List of animation FPS present in each scene of VTS_01_1.demuxed4.m2v
- 24fps foreground, 12fps background
- 24fps zoom out; most elements 12fps but helicopter rotors pop in & out @ 24fps
- 24fps zoom out; 12fps elements
- 24fps total; helicopter animated @ 12fps but cel is moved across background @ 24fps
- 24fps background; 12fps helicopter
- 24fps zoom in; static painting
- 24fps pan; slow dissolve appears to be at film rate but "total frame rate" in terms of unique fields is higher than 24fps because this edit was done on the telecined video rather than the film, and the two segments are out-of-phase with each other. Meaning scene 8 is on a different "pattern" or "cadence" than scene 9.
Last edited by Brad; 4th Apr 2020 at 01:19. Reason: Removed my incorrect conclusion regarding Forced Film; added entire post; added link regarding "on 2's" animation
Well, it seems that I left it at forced film in the field operation, while I was trying to find a solution yesterday. But now even when I put it to Honor flag, the video still comes at 23.976. I find it strange because when I safe the projet (without demuxing video) in DGindex the video is at 29.970 fps. Which is the VOB fps.
Anyway, I guest if I just demux the video with DGindex, no need to IVTV. But still remain the problem of stabilizing the video with QTGMC, those motion artifact are still present so are the jaggy lines if not using QTGMC.
I notice that with the forced film in DGindex, frame of the video goes back and forth, but with honor flag they go fluid forward, so here's an other sample.
I edited my post above. Adaptive IVTC is required for the '4' clip due to a video edit.
EDIT: I have no problem obtaining 29.970 from the samples in your first post. Not sure what issue you're having. The clip in post #5 is just a truncated version of the one in post #1 as far as I'm concerned...
EDIT2: Can you screenshot what you're referring to as jagged lines, with an arrow pointed to the offending area? (Or circle it.)
Last edited by Brad; 4th Apr 2020 at 01:25.
Well thanks vaporeon800, very apreciated man, but what do you mean by adaptative IVTC ?
And, do you have any idea about stabilizing those lines. Like I said QGTMC (in progressive mode) help a lot, but bring weird artifacts. Like lines don't follow through the motion and are one frame behind. Hard to explain, but you really noticed them when you have a close-up, they don't apear much on the background. Like if you look at eyes are mouth, you still can see the lines of the previous frame before.
Anyway big thanks again
Yes I'll send you a screen shot..
So for these motion artifact, if you look close to the mouth, there is the shadow of the previous frame blend into the next one. Like I said these are more obvious in a close-up, you can see the motion of eye, mouth, eye brown, of previous frames. Look at the mouth of the first sample. But that's when using QTGMC( Preset="Slow", InputType=1 )
[Attachment 52596 - Click to enlarge]
For the jaggy line look at the second sample. The problem seems to occurs when lines are to close from each others.
[Attachment 52597 - Click to enlarge]
Last edited by Altruo; 4th Apr 2020 at 01:47.
QGTMC ... bring weird artifacts. Like lines don't follow through the motion and are one frame behind. Hard to explain, but you really noticed them when you have a close-up, they don't apear much on the background. Like if you look at eyes are mouth, you still can see the lines of the previous frame before.
I'm wondering if you're really asking about the dot crawl and rainbows in the source.
Man you're really helpful. No I'm not refering to dot crawl and rainbows. I'm aware that rainblow (are something like that was present in the sample). Using something like DFMDeRainbow() resolved it in a way, but affect to much the color. It's more the jaggyness that I want to resolve since it's present on other part of the video or episodes without the rainbow. QTGMC resolve it a lot, but mess it up unfortunately with the motion trail has you called
EDIT: By jaggyness I mean mouvement among the lines (shakyness)
Sorry but I'm looking at the raw 29.970 version of the sample video, in the area your screenshot shows, and the only movement I see in that area is rainbowing.
Do you still see movement in that region with this script?
MPEG2Source("VTS_01_1.demuxed4.d2v") #Honor Pulldown Flags Greyscale()
Well they appear as well at the end of the sample within the building. But, it seems that I'm going to have to live with either that shakyness or the motion trail. Was just wandering if there was a way to fix it without going through all episode to find scene by scene to fix it with QGTMC.
But I admit that without the rainbow, doesn't look like much. I should just accept it...
By the way your very generous of your time...
TDeintted = QTGMC(FPSDivisor=2)
Last edited by manono; 4th Apr 2020 at 03:29.
But you said the jaggies were in the original 29.97 unaltered source:
Besides, these default settings don't trigger deinterlacing on the scene that you highlighted in yellow.
So I'm still confused, unless you looked at the unaltered source using a player rather than MPEG2Source (in that case your player deinterlaced it).
Do not use both TIVTC and QTGMC at the same time. Pick one or the other. Do not use QTGMC if you do not want those motion artifacts like on the mouth areas. That is a side effect of that filter that I have found no way around when it comes to animation sources.
Your samples are 23.976fps progressive and not interlaced. If your source is interlaced and still has some leftover combing artifacts you can lower visibility with this.
checkmate(thr=9, max=9, tthr2=0) # Will clean moving dot crawl artifacts. Use before deinterlacer. bifrost(scenelumathresh=4) # Should clean or reduce rainbows by quite a bit. Use before deinterlacer. (Use Bifrost v1.1. v2.0 is not as good at rainbow removal in my opinion) tfm(order=1).tdecimate(mode=1,hybrid=0) # Will deinterlace the video and make 23.976fps. santiag(strv=0, nns=4, nsize=5) # This will clean leftover pieces of combing if they are small. crop(0,0,0,0) # Crop by the numbers you want if you plan to crop anything. LanczosResize(640,480) # Resizes the video to 4:3 aspect ratio.
tfm(clip2=QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", FPSDivisor=2, sourcematch=3, tr1=2, tr2=2, NoiseTR=2, sharpness=0.1)).tdecimate(mode=1,hybrid=0)
If the source is progressive, then just use everything else without the deinterlacer.
Be sure to put santiag BEFORE you crop or resize, or the very bottom line row of pixels in your video will shift over some.
I would put all of these suggested filters before you crop or resize, and keep them in the order suggested.
For a tip about QTGMC, this will produce MUCH better quality output than the settings you posted in your post above. You will retain a lot more detail and lines will be more correctly in place.
QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", fpsdivisor=2, sourcematch=3, tr1=2, tr2=2, NoiseTR=2, sharpness=0.1)
Last edited by killerteengohan; 6th Apr 2020 at 19:14.