View Full Version : Questions regarding DeDup & QTGMC
ghostfromsparta
1st July 2020, 03:34
Hello everyone. This is my first post here although I've used these forums for information for a while. I'm only making this post because I haven't seen this topic covered and I need a little bit of help.
Thing is, I have an interlaced source, a DVD made from a mid-gen VHS copy of a concert, and my plan is to use QTGMC to deinterlace the footage (I've heard it's much better than yadif). Now, I'm also interested in removing the duplicated frames of the recording and interpolate it for it to look smoother. All of this through an Avisynth script.
Do I have to remove the dupped frames (DeDup) BEFORE deinterlacing or AFTER?
I'm not very experienced with this but my guess would be after, might be wrong.
How do I use DeDup to remove the dupped frames?
I ran a first pass (DupMC) on my footage and generated a log with the dupped frames, a large amount. But I don't know what to do now that I've got the log. I've tried to do the second pass (DeDup) but I get a series of errors, the most frequent is "DeDup: first pass missed frame 2". I truly have no clue on how to do it.
I'm very new to Avisynth too, but I have managed to get a few plugins working with no issue.
Is it good to interpolate to 60fps (original framerate) after the deinterlacing and the dup removal are done? Should I do it or just deinterlace the footage?
Those are my questions. Have a good one everyone.
johnmeyer
1st July 2020, 17:09
If you have dups in a capture that should not have any dups, and if the audio is in sync, then you almost certainly have an even bigger problem, namely dropped frames. The only way the video can have dups and also have audio that is in sync is if some frames are dropped. This situation is quite common when video is captured with a setup that is not properly adjusted. What happens is that the capture device cannot keep up with the incoming video so it drops a frame. In order to keep the video and audio in sync, it then adds a duplicate frame, usually a few frames later. You will generally find the dropped frame far more visually disturbing than the dups.
Both the dups AND the drops must be corrected before any temporal noise reduction is attempted and, quite frankly, I think you'll find the dropped frame artifact far more disturbing than the noise or the dups.
I posted a script in these forums which attempts to find the dropped frame and also the dups and attempts to correct both problems. However, to know whether such a script will work, you need to post 10-15 seconds of footage that is losslessly cut from your original.
Automatically fix dups followed (eventually) by drops (https://forum.doom9.org/showthread.php?t=161758)
ghostfromsparta
1st July 2020, 22:45
Thank you for your reply. I currently have no information on how the capture was made (it was a bootleg DVD released almost 20 years ago, I think). Audio and video are very much in sync so I'm not sure what to do. I will leave 15 seconds of footage so you can take a look. I'm sure the script you are talking about will be very helpful, specially because I'm trying to fix the recording as much as I can.
https://drive.google.com/file/d/1uHzRNF8tU4XiXEWEDjpyHqVTB2MNYfj5/view?usp=sharing
johnmeyer
2nd July 2020, 04:14
I don't have a codec on my computer that can read the video file, so I can't examine it closely. However, I was able to play it online. Unfortunately, because you chose a section that contains virtually no motion, I couldn't tell if it has drops or dups, or anything else. I certainly did not notice anything that was obvious.
I'm accustomed to trying to rescue some pretty bad video, but to be blunt, that video is so awful, I wouldn't waste ten seconds on it. My gosh, you can't even see his eyes, mouth or nose for most of the video!!!
My advice: try to improve the audio, as best you can, and forget about the video.
ghostfromsparta
2nd July 2020, 04:52
I don't have a codec on my computer that can read the video file, so I can't examine it closely. However, I was able to play it online. Unfortunately, because you chose a section that contains virtually no motion, I couldn't tell if it has drops or dups, or anything else. I certainly did not notice anything that was obvious.
I'm accustomed to trying to rescue some pretty bad video, but to be blunt, that video is so awful, I wouldn't waste ten seconds on it. My gosh, you can't even see his eyes, mouth or nose for most of the video!!!
My advice: try to improve the audio, as best you can, and forget about the video.
That's a bummer, well I couldn't tell dupped or drops either. Also, I'm writing a few scripts that have improved the visuals of the video a lot.
Here's a comparison: https://ibb.co/BVqJgjc
johnmeyer
2nd July 2020, 06:20
I see what you've done with your script. Here are a couple of things you should change, in order to get a better result.
1. Use some sort of videoscope or histogram so you can avoid ruining the black or white levels. I think you might have been trying to bring out some detail in his hair by brightening the image (which you did), but you ruined the black levels in the process. Most brightening should be done with a "gamma" control which affects only the midtones. You should not use a "brightness" control. Note how the background is now gray instead of near-black.
2. Make sure to pay attention to aspect ratio. You ended up squishing the video vertically, and your singer is now short and fat. This is easy to fix; you just need to pay attention.
3. Sharpening is a classic example of the old mantra, "less is more." If you apply too much sharpening, you can end up with a lot of artifacts. How much to use is definitely a matter of taste, and to my taste you are pretty close to the right amount, but you might consider a little less; certainly don't use more.
ghostfromsparta
6th July 2020, 05:08
I see what you've done with your script. Here are a couple of things you should change, in order to get a better result.
1. Use some sort of videoscope or histogram so you can avoid ruining the black or white levels. I think you might have been trying to bring out some detail in his hair by brightening the image (which you did), but you ruined the black levels in the process. Most brightening should be done with a "gamma" control which affects only the midtones. You should not use a "brightness" control. Note how the background is now gray instead of near-black.
2. Make sure to pay attention to aspect ratio. You ended up squishing the video vertically, and your singer is now short and fat. This is easy to fix; you just need to pay attention.
3. Sharpening is a classic example of the old mantra, "less is more." If you apply too much sharpening, you can end up with a lot of artifacts. How much to use is definitely a matter of taste, and to my taste you are pretty close to the right amount, but you might consider a little less; certainly don't use more.
I actually used a VirtualDub script (a bunch of Ghost Removers) to remove the abberration present on the picture, it turns out it brought back a lot of detail, and as you said, it ruined the black levels. I already have another script for the aspect ratio, although I haven't sharpened (yet).
I will keep working on the video, thank you very much for your advice.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.