VideoHelp Forum
+ Reply to Thread
Results 1 to 18 of 18
Thread
  1. Member spiritgumm's Avatar
    Join Date: Jan 2004
    Location: United States
    Search Comp PM
    having a problem joining interlaced dvd video after processing.
    I have couple dvd videos with very short segments that I want to delogo. Rather than process the entire video, I cut out the segment with Videoredo. Dgindex the segment for Avisynth, deinterlace, delogo, re-encode with HC ENC (in Interlace mode), and insert segment back with Videoredo.
    When viewed on MPC-HC, the edit points show a brief black screen. I have MPC player output set to Directshow Video-> System Default. It plays fine if I switch that to Enhanced Renderer. While it might be the renderer, there must be something triggering the problem. As it might it be the interlacing, I'm posting my script below.
    MPEG2Source("C:\ord cut\VTS_01_1.d2v", cpu=0)
    assumetff()
    yadif(mode=1)
    a=trim(0,221).XLogo("LogoFinal2.bmp", X=98, Y=42, alpha=2)
    b=trim(222,0)
    a+b
    ChangeFPS(59.94)
    SeparateFields()
    SelectEvery(4,0,3)
    Weave()

    I also thought it might be a colorspace issue, so running the Before and After clips in Avisynth with Info showed the original was YV12, the processed clip in YU12. MediaInfo says they're both YUV.
    My video edit occurs at scene change: edit test.m2v
    Attached Files
    Quote Quote  
  2. I see no black frame viewing the M2V in MPC-BE or in VDubMod, or in VDub using an AviSynth script.

    Also, movies, by definition, aren't interlaced. You should unblend it:

    Yadif(Mode=1)#or QTGMC
    Srestore(frate=23.976)


    Rather than process the entire video, I cut out the segment with Videoredo...
    I do this sort of thing using ReplaceFramesSimple, part of stickboy's ReMapFrames.dll. No cutting and splicing. Maybe your problem is due to the cutting of interlaced material, not sure.

    Last edited by manono; 12th Jan 2014 at 22:54.
    Quote Quote  
  3. Member spiritgumm's Avatar
    Join Date: Jan 2004
    Location: United States
    Search Comp PM
    the video on both dvds (which are unrelated) looks to have originated as amateur or semi-pro film transfers to VHS - would that explain the interlacing?
    Would ReplaceFramesSimple work while providing some kind of passthrough so the majority of video was not re-encoded? I'd rather not re-encode the entire dvds; I only need to fix 4 seconds of a VCR clock on one, and 4 minutes of a company logo on the other.

    Never heard of MPC-BE. Not sure I like it compared to MPC-HC. With Video output set to System Default or some other settings, I see thick black lines at the scene change. In other settings, the video is choppy. Maybe I need to add some codecs...although again, I'd prefer the video not elicit any errors.
    Here's the complete "Before" clip for the edited "After" video posted above.
    VTS_01_1.demuxed.m2v
    Attached Files
    Quote Quote  
  4. Originally Posted by spiritgumm View Post
    Would ReplaceFramesSimple work while providing some kind of passthrough so the majority of video was not re-encoded?
    No. AviSynth works only on uncompressed frames. So everything that is processed by it must be re-encoded (or saved as uncompressed). With MPEG 2 video you manually "smart" edit by using a lossless MPEG 2 editor like Mpg2Cut2, cutting the video into sections (on i-frames), reencoding the sections you need to process, then joining all the section back together again at the end.
    Quote Quote  
  5. Originally Posted by spiritgumm View Post
    Would ReplaceFramesSimple work while providing some kind of passthrough so the majority of video was not re-encoded?
    No.
    I'd rather not re-encode the entire dvds
    Then what was the point of posting a script? If used, you're reencoding. That nasty sample should be reencoded anyway, if only to get rid of the blending. It also has to be greyscaled and denoised and a bunch of other stuff, depending on how much time you want to devote to it. It can be significantly improved with minimum work.
    With Video output set to System Default or some other settings, I see thick black lines at the scene change.
    Geez, then turn on the deinterlacing. The whole sample is full of interlacing and I prefer to see it as it gives a clue as to what needs doing. It's showing you exactly how it's stored on the DVD. I agree it looks bad and think it should be gotten rid of.
    Here's the complete "Before" clip for the edited "After" video posted above.
    That one isn't blended, but only needs an IVTC to return it to progressive 23.976fps. Only now I realize that the part before the scene change in your first sample isn't blended but hard telecined. Not sure why you didn't use that version of the film all the way through as an IVTC is way better than unblending. Maybe the hard telecined version has serious problems of its own?
    Last edited by manono; 13th Jan 2014 at 14:21.
    Quote Quote  
  6. Member spiritgumm's Avatar
    Join Date: Jan 2004
    Location: United States
    Search Comp PM
    Originally Posted by jagabo View Post
    Originally Posted by spiritgumm View Post
    Would ReplaceFramesSimple work while providing some kind of passthrough so the majority of video was not re-encoded?
    No. AviSynth works only on uncompressed frames. So everything that is processed by it must be re-encoded (or saved as uncompressed). With MPEG 2 video you manually "smart" edit by using a lossless MPEG 2 editor like Mpg2Cut2, cutting the video into sections (on i-frames), reencoding the sections you need to process, then joining all the section back together again at the end.
    that's what I thought, but since I originally stated I did not want to re-encode entire dvd, I thought I was missing something in manono's advice.
    I did try mpg2cut2, but not in the order you suggested, so I should try it again, but chances are the error is from my mismatching video. I mistook the video as interlaced because it doesnt behave like typical telecine. If you check out the original "Before" video, the field pattern is 2.2.1, and when I IVTCed, the frames are choppy - the car turning the corner isnt smooth; seems to clear up afterwards. If it's deblended with yadif/srestore, the car moves smoothly...but if I want it to match the original video, should I omit any frame-splitting, and just delogo it?

    Regarding the videos posted, the first was just to show the edit point issue. The script-processed video comes first, and at scene change is followed by original video.
    The 2nd video is the segment before the scene change, but unprocessed.
    Quote Quote  
  7. Originally Posted by spiritgumm View Post
    that's what I thought, but since I originally stated I did not want to re-encode entire dvd, I thought I was missing something in manono's advice.
    You posted a script. Using AviSynth means reencoding. Therefore what you wrote in that first post was contradictory. In addition, both sources are so bad they should be reencoded.
    If you check out the original "Before" video, the field pattern is 2.2.1, and when I IVTCed, the frames are choppy
    You're right. As the car comes into the picture and turns the corner, just before and during the time the 'BUS STOP' text comes on-screen, it does play jerky. TIVTC messes up, probably because it's too dark and the movement too far away. Try this, something jagabo cooked up for something else:

    Interleave(TFM(field=1), TFM(field=0))
    Srestore(frate=23.976)


    TFM is doing the field matching and Srestore the decimation. Sometimes with tape sources things don't behave normally.
    ...should I omit any frame-splitting, and just delogo it?
    I wouldn't think of splitting off a section for special treatment if I could avoid it. Is there a reason you're mixing and matching your sources?
    Quote Quote  
  8. Originally Posted by spiritgumm View Post
    If you check out the original "Before" video, the field pattern is 2.2.1, and when I IVTCed, the frames are choppy - the car turning the corner isnt smooth; seems to clear up afterwards. If it's deblended with yadif/srestore, the car moves smoothly...
    Yes, you would use yadif/srestore to return it to 25 fps progressive, not TFM()/TDecimate().

    Originally Posted by spiritgumm View Post
    but if I want it to match the original video, should I omit any frame-splitting, and just delogo it?
    I guess you mean, should you keep the original MPEG 2 video for the parts that don't need delogoing. That's up to you. It depends on whether you think you can clean up the parts without logos enough to make it worth reecoding, or not.

    Originally Posted by spiritgumm View Post
    Regarding the videos posted, the first was just to show the edit point issue. The script-processed video comes first, and at scene change is followed by original video.
    The 2nd video is the segment before the scene change, but unprocessed.
    I didn't see anything wrong with the transition. But the frame at the transition is interlaced with one field from the first clip (a blended field) and another from the second. Maybe your graphics card is trying to deinterlace and screwing up. It plays fine here.
    Quote Quote  
  9. Member spiritgumm's Avatar
    Join Date: Jan 2004
    Location: United States
    Search Comp PM
    manono, I specified at the start I was only delogoing and re-encoding a small segment.
    Suffice it to say I dont want to deal with fixing all the issues with the dvds. It takes me longer (than you and other members here) to figure out what to do, how to do it, hunt down the filters, make tests, get more advice, etc. I just wanted to fix these glaring issues because they are brief. The dvds can always be improved in the future, and hopefully the brief segments wont look too bad being re-encoded yet again.

    When I wrote 'omit frame-splitting and just delogo,' I was referring to field separation in the script: should I omit any IVTC or deinterlacing steps in order to output the same crappy video blends so it matches the original?

    Btw, is there an explanation for the 2.2.1 pattern?
    If there are blended frames (and if I were going to process the entire dvd, which I'm not), why not use yadif/srestore?
    Quote Quote  
  10. Member spiritgumm's Avatar
    Join Date: Jan 2004
    Location: United States
    Search Comp PM
    Originally Posted by jagabo View Post
    Originally Posted by spiritgumm View Post
    If you check out the original "Before" video, the field pattern is 2.2.1, and when I IVTCed, the frames are choppy - the car turning the corner isnt smooth; seems to clear up afterwards. If it's deblended with yadif/srestore, the car moves smoothly...
    Yes, you would use yadif/srestore to return it to 25 fps progressive, not TFM()/TDecimate().
    it should be 24 fps, I hope.
    Originally Posted by jagabo View Post
    Originally Posted by spiritgumm View Post
    but if I want it to match the original video, should I omit any frame-splitting, and just delogo it?
    I guess you mean, should you keep the original MPEG 2 video for the parts that don't need delogoing. That's up to you. It depends on whether you think you can clean up the parts without logos enough to make it worth reecoding, or not.
    what I mean is delogoing the "Before" clip without separating the video into fields, thus outputting the same blends or whatever is going on:

    MPEG2Source("C:\ord cut\VTS_01_1.d2v", cpu=0)
    a=trim(0,221).XLogo("LogoFinal2.bmp", X=98, Y=42, alpha=2)
    b=trim(222,0)
    a+b

    The only reason I used a 30 second clip when I only need to delogo 4 seconds was to place the edit at a scene change.
    Quote Quote  
  11. Originally Posted by spiritgumm View Post
    Originally Posted by jagabo View Post
    Yes, you would use yadif/srestore to return it to 25 fps progressive, not TFM()/TDecimate().
    it should be 24 fps, I hope.
    No, 25 fps. It looks like the usual PAL to NTSC conversion. You can slow it down to 23.976 fps with AssumeFPS(24000,1001) but then you'll have to slow the audio too. But I didn't look too closely, maybe it should go directly to 23.976 with SRestore().

    Originally Posted by spiritgumm View Post
    what I mean is delogoing the "Before" clip without separating the video into fields
    I would revert back to clear film frames.

    <edit>

    After taking a closer look, SRestore(frate=23.976, thresh=12) seems to generate smoother video than SRestore(frate=25, thresh=12). I'd check a few other shots though.
    Last edited by jagabo; 13th Jan 2014 at 21:25.
    Quote Quote  
  12. Member spiritgumm's Avatar
    Join Date: Jan 2004
    Location: United States
    Search Comp PM
    I cant quite tell from the actor's voices (which I'm familiar with) if they're slightly higher pitched. I'm just surprised some 16mm USA films would have been transferred in UK or wherever.
    Since I'm ONLY trying to fix the VCR clock (and logo on other video), and want to insert the clip back into the source, I wont be converting to 24 or 25 fps.
    Quote Quote  
  13. Originally Posted by spiritgumm View Post
    When I wrote 'omit frame-splitting and just delogo,' I was referring to field separation in the script: should I omit any IVTC or deinterlacing steps in order to output the same crappy video blends so it matches the original?
    Sorry, but if I start a project I don't do only a little bit of what should be done. I'd unblend and make the whole thing progressive again, remove the logos, and other things as well.
    I'm just surprised some 16mm USA films would have been transferred in UK or wherever.
    Maybe some, but not this one.
    Quote Quote  
  14. Member spiritgumm's Avatar
    Join Date: Jan 2004
    Location: United States
    Search Comp PM
    So after fixing the VCR clock (and logo on other video), I want to insert the clip back into the source. What's the best way to output video that will be compatible with original?
    I tried my script in post #10, encoding progressive, but playing it with MPC-HC shows combing. Encoding interlaced, no combing in playback, and resembled the original video. I thought it would have been the reverse, so I'm confused.
    Quote Quote  
  15. If your video has interlaced frames you must encode interlaced. Encoding progressive doesn't convert interlaced frames to progressive frames. It just controls how the encoder handles the frames internally and what it tells the player when the video is played.
    Quote Quote  
  16. Member spiritgumm's Avatar
    Join Date: Jan 2004
    Location: United States
    Search Comp PM
    what is the original video (in post #3)? Is it hard telecine? blended? interlaced?
    Quote Quote  
  17. Originally Posted by spiritgumm View Post
    what is the original video (in post #3)? Is it hard telecine? blended? interlaced?
    It's hard telecined with blended fields, encoded interlaced.
    Quote Quote  
  18. Member spiritgumm's Avatar
    Join Date: Jan 2004
    Location: United States
    Search Comp PM
    well, I'm not having any luck trying various scenarios.
    Videoredo is GOP-accurate, and seems to cut and combine better than mpg2cut2. It combines random clips from the video seamlessly, so it's some incompatibility with the processed video.
    Quote Quote  



Similar Threads