VideoHelp Forum
+ Reply to Thread
Results 1 to 27 of 27
Thread
  1. This has probably been asked, but I haven't the slightest idea what terms to search for to find my answer. I've looked up everything I could think of.

    I have an old bootleg (unfortunately) anime I'm trying to rip and improve the video. I've managed to get VirtualDub and find a few plugins (a feat for me coz I'm pretty Windows/video stupid here).

    The "noise" is present during playback even after the filters and even in the lighter scenes. It's like static on an old tv or something? I'm not sure how to describe it. Is there a way to smooth it out without losing too many details? Any suggested filters I'm unaware of that will work better?

    Here are a few shots from VirtualDub as examples of what I have so far:
    Click image for larger version

Name:	Screen Shot 2014-11-05 at 4.04.40 PM.png
Views:	303
Size:	906.4 KB
ID:	28396
    Click image for larger version

Name:	Screen Shot 2014-11-05 at 4.04.09 PM.png
Views:	421
Size:	838.8 KB
ID:	28398

    And the filters I'm using:
    Click image for larger version

Name:	Screen Shot 2014-11-05 at 3.57.46 PM.png
Views:	1814
Size:	10.3 KB
ID:	28397

    If you need a video sample what length, format, etc. would it need to be?

    Thanks in advance for any advice.
    Quote Quote  
  2. Originally Posted by nobodyhome View Post
    If you need a video sample what length, format, etc. would it need to be?
    In the 'source' format. 10m seconds or so with steady movement would be plenty. And anime isn't deinterlaced, using either blend mode or anything else. It's usually IVTC'd. For the best results, if you're serious about this restoration effort you'll need to learn AviSynth.
    Quote Quote  
  3. I'll look into AviSynth, but I'm pretty bad at this sort of thing. I'll certainly give it a try.

    Any starting points or any suggestions for VirtualDub though?

    The source is a dvd rip so the videos are in vob(?). I'm not sure how to cut that up, but I used VirtualDub to clip a couple parts.
    One during a dark portion and one during a lighter portion.

    https://mega.co.nz/#!dQxBkbLa!cCqxqC7DaWdRk5zbBN8MjsBEUwZeiK5JvwRbjYAHFNM
    https://mega.co.nz/#!1NRihSCK!MDWAdPyy98MDpXuG5XNsMLZlQDskrYuV8DvUOQFdSrQ

    I don't even know if I'm doing this right. I want to learn though. Thx.
    Quote Quote  
  4. Banned
    Join Date
    Oct 2014
    Location
    Northern California
    Search PM
    Originally Posted by nobodyhome View Post
    And the filters I'm using:
    Image
    [Attachment 28397 - Click to enlarge]


    If you need a video sample what length, format, etc. would it need to be?

    Thanks in advance for any advice.
    Definitely provide a sample but in the mean time why the tiny resize? Nothing good comes from those tiny resizes.

    I like to apply an edge aware blur for animations and then apply a "Lumetri" or some "film stock" kind of look.
    Last edited by newpball; 5th Nov 2014 at 22:33.
    Quote Quote  
  5. Originally Posted by newpball View Post
    Originally Posted by nobodyhome View Post
    And the filters I'm using:
    Image
    [Attachment 28397 - Click to enlarge]


    If you need a video sample what length, format, etc. would it need to be?

    Thanks in advance for any advice.
    Definitely provide a sample but in the mean time why the tiny resize? Nothing good comes from those tiny resizes.

    I like to apply an edge aware blur for animations.
    This probably comes from my severe lack of knowledge, but I read somewhere that the standard max size for dvd video was 720 x 480 and that a 4:3 ratio was 640 x 480 where black boarders are added to make it 720 x 480. The original video from the bootleg is distorted so I was trying to make it back to a 4:3 size.

    Again, I'm probably getting something horribly wrong which is why I'm asking here. I'm very new to this and have hit a wall, so to speak. I'm more than open to guidance.

    Ideally I want to re-burn the improved video to dvd to pseudo-replace my bootleg dvd as I can't buy the original.

    EDIT: "edge aware blur"? I'll have to google that. What settings do you suggest? Also, is that in VirtualDub or AviSynth?
    Quote Quote  
  6. Originally Posted by nobodyhome View Post
    The source is a dvd rip so the videos are in vob(?)
    Open a VOB in DGIndex, use the '[' and ']' buttons to mark a start and end point. Go File->Save Project and Demux Video. Upload the resulting M2V.
    ...but I used VirtualDub to clip a couple parts.
    Useless. 546MB? Like I said, 10 seconds or so of a scene with steady movement.
    ...but I read somewhere that the standard max size for dvd video was 720 x 480 and that a 4:3 ratio was 640 x 480 where black boarders are added to make it 720 x 480.
    You have it all garbled.
    Quote Quote  
  7. Originally Posted by manono View Post
    Useless. 546MB? Like I said, 10 seconds or so of a scene with steady movement.
    Hopefully this is better?
    https://mega.co.nz/#!MVxz2YZS!tx_DltxwaAtdK-9LoAJZmKyzgsNZknDFeeal3XsdTTY

    You have it all garbled.
    Is it wrong or just garbled? I'm trying to make sure I understand here.
    Quote Quote  
  8. Member
    Join Date
    Sep 2012
    Location
    Australia
    Search Comp PM
    The majority of DVDs are 720x[480/576] but PAL/NTSC CRT display pixels aren't square, so to fit a modern square pixel display they need to be resized to something like 640x480 for 4:3 NTSC, 768x576 for 4:3 PAL (it doesn't really matter how you resize them, as long as they're still 4:3 by the end), 853(854)x 480 for 16:9 NTSC or 1024x576 for 16:9 PAL. Some DVDs are 704, literally, but others are 704 erroneously encoded as 720 with black bars, for the 704 as 720 it's best to crop a total of 16 pixels from the sides (usually 8 and 8) and then resize to square pixels. Of course, since non-square pixels are pretty standard fare, any player worth it's salt can do the resizing on playback, there are elements in the video streams that tell the player how to resize the video to the correct dimensions and some people prefer to use that instead of resizing the video themselves (resizing can be considered damaging to the image quality through altering the picture).
    Quote Quote  
  9. Originally Posted by nobodyhome View Post
    Hopefully this is better?
    Much better, thanks. And smaller, too.

    Well, you have a terrible field-blended video. Its true framerate is supposed to be 23.976fps. If you deinterlace as you showed in VDub you'll get 29.97fps and a blended blurry mess. The only way to handle it properly is with an AviSynth bobber and unblender. Nothing else can even come close. A sample script to unblend it and return the correct framerate might be:

    QTGMC()
    SRestore(Frate=23.976)


    Your original question about denoising it is secondary to unblending it and getting the right framerate, in my opinion. If you really want, as you said in your first post, to 'improve' it, then you have no choice but to use AviSynth with it.
    Quote Quote  
  10. Banned
    Join Date
    Oct 2014
    Location
    Northern California
    Search PM
    Originally Posted by manono View Post
    Originally Posted by nobodyhome View Post
    Hopefully this is better?
    If you deinterlace as you showed in VDub you'll get 29.97fps and a blended blurry mess.
    Really?

    Deinterlacing the video seems to work just fine for me (including the blended frames). It seems to me that after deinterlacing he needs to apply some smart decimation (that detects blended frames), the footage is short and at first sight a decimation of 1.25 should do the job but after frame 44 that seems to go awry.

    Apart from that the video levels are not right and the colors could use some upgrading.
    Quote Quote  
  11. Originally Posted by newpball View Post
    Originally Posted by manono View Post
    Originally Posted by nobodyhome View Post
    Hopefully this is better?
    If you deinterlace as you showed in VDub you'll get 29.97fps and a blended blurry mess.
    Really?

    Deinterlacing the video seems to work just fine for me (including the blended frames). It seems to me that after deinterlacing he needs to apply some smart decimation (that detects blended frames), the footage is short and at first sight a decimation of 1.25 should do the job but after frame 44 that seems to go awry.

    Apart from that the video levels are not right and the colors could use some upgrading.
    That's what manono is referring to. That script detects the blends, preferentially throws them away, giving you the original 23.976 frames. It's not perfect but it might miss one or two, or maybe some sources have too many blends, but it's a far superior method.

    You can decimate in vdub, but it's not adaptive. It always picks every "nth" frame. But if cadence changes, pattern changes, you're in trouble because you might throw away the "wrong" good frame. In avisynth , adaptive decimation algorithms can be used that can change when different characteristics are encountered

    Yes levels need to be adjusted
    Quote Quote  
  12. Originally Posted by newpball View Post
    Really?
    Yes, really.
    Quote Quote  
  13. Banned
    Join Date
    Oct 2014
    Location
    Northern California
    Search PM
    Originally Posted by manono View Post
    Originally Posted by newpball View Post
    Really?
    Yes, really.
    Well you are wrong. Did you actually try it?

    I deinterlaced the video with yadiff double framerate and every frame is satisfactory deinterlaced (even the blended ones ). After that you can smart remove the blended frames using what looks like a 1.25 decimate. The deinterlacing and the blending have nothing to do with each other in this clip.
    Last edited by newpball; 6th Nov 2014 at 15:08.
    Quote Quote  
  14. Originally Posted by newpball View Post
    Originally Posted by manono View Post
    Originally Posted by newpball View Post
    Really?
    Yes, really.
    Well you are wrong. Did you actually try it?

    I deinterlaced the video with yadiff double framerate and every frame is satisfactory deinterlaced (even the blended ones ). After that you can smart remove the blended frames using what looks like a 1.25 decimate.


    It's ok to disagree - that's what forum discussion is for

    Vdub certainly an option, and the OP wanted to use vdub. Easy to use, case closed right?

    But maybe the better question is did you try the avisynth method?

    Maybe vdub method is "good enough" for his purposes , or he can't be bothered to use avisynth

    But vdub method produces worse results, without a doubt. Easy to demonstrate and prove this. Having fewer to no blurry frames is better than having blurry frames. Proven higher quality deinterlacer is better than lower quality deinterlacer that leaves deinterlacing artifacts. Not only that, vdub doesn't handle interlaced 4:2:0 chroma properly and can introduce new chroma artifacts

    BMW 335i is a nice car , until you drive a Porsche 918
    Quote Quote  
  15. Banned
    Join Date
    Oct 2014
    Location
    Northern California
    Search PM
    Originally Posted by poisondeathray View Post
    Vdub certainly an option, and the OP wanted to use vdub. Easy to use, case closed right?
    I think that Vdub can certainly do the deinterlacing and frame doubling adequately. And actually the poster may stop at that level. The blending for this animation is not going to be very noticeable on normal viewing speeds.

    But of course if you want to go a step beyond you need to smart decimate. As I already wrote the 1 in 5 cadence is interrupted around frame 44 so a fixed decimate won't work.

    Originally Posted by poisondeathray View Post
    But maybe the better question is did you try the avisynth method?
    I tried and got "invalid response to CACHE_GETCHILD_CACHE_MODE...." from Srestore()". I suppose that is because I am playing with some of the MT versions for QTGMC()

    Originally Posted by poisondeathray View Post
    Maybe vdub method is "good enough" for his purposes , or he can't be bothered to use avisynth
    Perhaps, it seems to me that a straightforward deinterlacing with frame doubling improves this already a lot, that with adjusting the levels and giving the colors a little boost and perhaps some light edge aware surface blur should improve the video a lot compared to where we started from.


    Originally Posted by poisondeathray View Post
    But vdub method produces worse results, without a doubt. Easy to demonstrate and prove this. Having fewer to no blurry frames is better than having blurry frames.
    No argument here.

    I do not think there is much of an argument as long as everybody agrees that the interlacing and frame blending (not field blending!) are two separate and independent issues.
    Quote Quote  
  16. Originally Posted by newpball View Post
    I tried and got "invalid response to CACHE_GETCHILD_CACHE_MODE...." from Srestore()". I suppose that is because I am playing with some of the MT versions for QTGMC()
    I haven't seen this error message before, but MT and SRestore() don't mix. You're better off using single threaded mode



    I do not think there is much of an argument as long as everybody agrees that the interlacing and frame blending (not field blending!) are two separate and independent issues.
    Actually deinterlacing and blending have everything to do with this clip - that' s how this video got in the sorry state in the first place: blend deinterlaced norm conversion

    If you're talking about the processing of the clip now - then yes because the blends were already present in the fields (it's field blended, not frame blended). So how you deinterlace doesn't really affect the blends because they are already there (well it can affect it, if you use a temporal deinterlacer, or blend deinterlacer)
    Quote Quote  
  17. "invalid response to CACHE_GETCHILD_CACHE_MODE....
    Yup, I've had this many times in the past few months. You can find the solution over at doom9.org, but to save you the search, you need to make sure you are using the correct version of mt_masktools-25.dll (or mt_masktools-26.dll). In each case where I've had this error, that was the problem,
    Quote Quote  
  18. Originally Posted by newpball View Post
    I tried and got "invalid response to CACHE_GETCHILD_CACHE_MODE...." from Srestore()".
    Add SetMtMode() at the start of your script. Specify mode and threads if necessary.
    Quote Quote  
  19. Originally Posted by newpball View Post
    Well you are wrong. Did you actually try it?
    Why try something I already know produces inferior results? Besides, in my original comment I mentioned the way he did it (the VDub blend deinterlacer) was the wrong thing to use. One doesn't single-rate deinterlace at all when you have blended fields. One uses an unblending filter following a good bobber. As pdr mentioned, it's not perfect, but it's way better than anything else. That and you can use the much better AviSynth bobbers.
    I deinterlaced the video with yadiff double framerate and every frame is satisfactory deinterlaced (even the blended ones )
    Which is not how the OP did it and not how you implied you did it when you said first, "Deinterlacing the video seems to work just fine for me (including the blended frames)." I'm not denying you might get better results doing it your way as compared to his. I'm saying you'll get the best results bobbing and unblending through AviSynth. If one really wants to work with this kind of material, why not learn to do it right?
    Quote Quote  
  20. Banned
    Join Date
    Oct 2014
    Location
    Northern California
    Search PM
    Thanks, I ripped the MT stuff out and I got this to work now!
    Quote Quote  
  21. SRestore() works fine with MT.
    Quote Quote  
  22. Srestore can work with MT, but you have a higher probability of mixed up frames . It's not worth the risk IMO

    @newpball - when using filters like srestore, remember not to scrub around the timeline. They only work linearly or if you process a few frames before the region of interest
    Quote Quote  
  23. Originally Posted by poisondeathray View Post
    when using filters like srestore, remember not to scrub around the timeline.
    Yes. And combining SRestore with other out-of-order filters can be an problem.
    Quote Quote  
  24. Wow. As everybody spun off into jargon I barely understand I finally sat down and started teaching myself Avisynth. I think I created my first script, but haven't gone much further than that. I'm reading all the posts here and trying my best to understand, but unfortunately it's still Greek to me. I'm a horrible beginner here.

    Still trying to figure out what the script would even look like and how to make it all work. Thanks though for seriously helping me with this. Sorry I don't understand more quickly than this.
    Quote Quote  
  25. Banned
    Join Date
    Oct 2014
    Location
    Northern California
    Search PM
    Originally Posted by nobodyhome View Post
    Wow. As everybody spun off into jargon I barely understand I finally sat down and started teaching myself Avisynth. I think I created my first script, but haven't gone much further than that. I'm reading all the posts here and trying my best to understand, but unfortunately it's still Greek to me. I'm a horrible beginner here.

    Still trying to figure out what the script would even look like and how to make it all work. Thanks though for seriously helping me with this. Sorry I don't understand more quickly than this.
    Actually almost anything you need to know is already said here.

    1. Download the 'Alpha' 2.6 version of Avisynth, which will install in a folder called Avisynth 2.5 (don't ask why )
    2. Download QTMG plugin package for 2.6 with all requirements (no MT versions if you are just starting) and follow the instructions.
    3. Download the SRestore function and save it in the plugins folder under Avisynth 2.5
    4. Create a script, call it myscript.avs or whatever you want in your favorite text editor
    5. Put in the script:

    MPEG2Source(" - location of your d2v file - ")
    QTGMC()
    SRestore(Frate=23.976)

    6. Save it.
    7. Double click on the script and make VirtualDub the default application to open this file.
    8. Then you are in Virtualdub and save as an AVI with you favorite compression, also don't forget to check the right color spaces both in and out.

    That should get you started!

    By the way you would lose the sound but you can merge it from the original in your favorite NLE or add more instructions to the script.

    Last edited by newpball; 6th Nov 2014 at 20:25.
    Quote Quote  
  26. Okay! I think I'm (slowly) getting the hang of this. I finally got Avisynth installed and VirtualDub to load the file without giving me a ton of errors (I'm telling you. I'm Windows stupid).

    I made the script as you instructed, used DGIndex to make the .d2v file (same process that I used to make the sample), and got everything to load.

    From here if I want to cropping, denoise, or anything else would it be more effective to use more Avisynth scripts or do I use VirtualDub? If I use scripts what scripts should I start looking at? I have no idea what anything is called when it comes to video.

    Thanks again and again for helping me get a good foothold on all this.
    Quote Quote  
  27. Member
    Join Date
    Sep 2012
    Location
    Australia
    Search Comp PM
    Originally Posted by newpball View Post
    1. Download the 'Alpha' 2.6 version of Avisynth, which will install in a folder called Avisynth 2.5 (don't ask why )
    Because it's an Alpha, there's no such thing as Avisynth 2.6 yet because it's not actually finished, what you've downloaded and installed is nothing more than a modified version of Avisynth 2.5. When the final version of 2.6 is released I'm pretty sure it WILL install into a folder called Avisynth 2.6, as will any 2.7 alphas that follow.

    -Edit- I'm not sure why he wasn't supposed to ask, or why you didn't just tell him...
    Quote Quote  
Visit our sponsor! Try DVDFab and backup Blu-rays!