VideoHelp Forum


Try DVDFab Video Downloader and rip Netflix video! Or Try DVDFab and copy Blu-rays! or rip iTunes movies!
+ Reply to Thread
Page 2 of 2
FirstFirst 1 2
Results 31 to 52 of 52
Thread
  1. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    I know. I just restored my AVI's onto one drive and tried it. Works OK. I might have tried it earlier with a typo there. My fault (again).
    Last edited by sanlyn; 21st Mar 2014 at 08:30.
    Quote Quote  
  2. I have used Jmac's suggestion , using pointresize solve the chroma issue it appears.
    The current rgb>yv12 scheme is bugged to say the least.

    pointresize(last.width, last.height*4).converttoyv12(matrix="Rec601",inter laced=true).pointresize(last.width, last.height/1)

    The full script:

    MPEG2Source("F:\00-videos_clips\MyVideo.d2v", cpu=0,info=3)

    ## DEBLOCKING INTERLACED SOURCE :
    par=getparity()
    SeparateFields().PointResize(width,height)
    Deblock_QED(24,28,uv=3)
    AssumeFrameBased()
    SeparateFields()
    Merge(SelectEven(),SelectOdd())
    par ? AssumeTFF() : AssumeBFF()
    Weave()
    #
    ConvertToRGB32(matrix="Rec601",interlaced=true)
    LoadVirtualDubPlugin("C:\Program Files (x86)\virtualdub1.9\plugins\NeatVideo.vdf", "NeatVideo", 2)
    NeatVideo("C:\Program Files (x86)\Neat Video for VirtualDub\PROFILES\76%-deblock_entrelacé-NEW.dnp", "C:\Program Files (x86)\Neat Video for VirtualDub\PRESETS\profil_76%.nfp", "1.0", "2", "1", "1" )
    #
    pointresize(last.width, last.height*4).converttoyv12(matrix="Rec601",inter laced=true).pointresize(last.width, last.height/1)
    Quote Quote  
  3. Are you using avisynth 2.6.x or 2.5.x ? Because there are improvements to the way chroma is handled
    Quote Quote  
  4. The current rgb>yv12 scheme is bugged to say the least.
    So you're saying the issue is with how avisynth handles (interlaced) RGB=>YV12 ?

    But if you comment out the last line, is it still fubar ? (i.e leave it in RGB)
    Quote Quote  
  5. If i leave it in rgb32 it looks just fine to me
    Quote Quote  
  6. Weird. Why does it only affect Neat video processed video? It doesn't occur with other interlaced content

    Gavino will probably figure it out
    Quote Quote  
  7. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    I see at the avisynth main page that 2.6 is still in alpha. I'm sticking with 2.5.8 for now. Anyone having problems with 2.6? I'm willing to give it a try if no major issues.
    Last edited by sanlyn; 21st Mar 2014 at 08:31.
    Quote Quote  
  8. No big problems with 2.6 here. I've been using it for about a year.

    There are minor bugs, but they have been reported

    Notably, it fixes the interlaced chroma shift bug in 2.5.8 Gavino noted here ( http://forum.doom9.org/showthread.php?t=147629 )and a few other fixes. It adds other colorspaces (for example YV24 or Y'CbCr 4:4:4), other chromasampling options
    Quote Quote  
  9. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    As usual, poisondeathray, I'm sold. Stopping everything now to get 2.6. Thanks.
    Last edited by sanlyn; 21st Mar 2014 at 08:31.
    Quote Quote  
  10. If you're switching to 2.6, beware that some plugins and filters have specific versions for avisynth 2.5.x or 2.6.x

    For example masktools is like this. Make sure you only have the correct corresponding version (remove the other dlls for 2.5.x)
    Quote Quote  
  11. @themaster1 - Can you upload a lagarith version a few frames of everything up to the last line for Gavino to examine (I don't think he uses neat video)

    So basically this

    MPEG2Source("F:\00-videos_clips\MyVideo.d2v", cpu=0,info=3)

    ## DEBLOCKING INTERLACED SOURCE :
    par=getparity()
    SeparateFields().PointResize(width,height)
    Deblock_QED(24,28,uv=3)
    AssumeFrameBased()
    SeparateFields()
    Merge(SelectEven(),SelectOdd())
    par ? AssumeTFF() : AssumeBFF()
    Weave()
    #
    ConvertToRGB32(matrix="Rec601",interlaced=true)
    LoadVirtualDubPlugin("C:\Program Files (x86)\virtualdub1.9\plugins\NeatVideo.vdf", "NeatVideo", 2)
    NeatVideo("C:\Program Files (x86)\Neat Video for VirtualDub\PROFILES\76%-deblock_entrelacé-NEW.dnp", "C:\Program Files (x86)\Neat Video for VirtualDub\PRESETS\profil_76%.nfp", "1.0", "2", "1", "1" )
    So you are saying everything is fine upto there (still in RGB32), but as soon as you put ConvertToYV12(matrix="Rec601", interlaced=true), you get problems ?
    Quote Quote  
  12. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    Originally Posted by poisondeathray View Post
    If you're switching to 2.6, beware that some plugins and filters have specific versions for avisynth 2.5.x or 2.6.x

    For example masktools is like this. Make sure you only have the correct corresponding version (remove the other dlls for 2.5.x)
    My plugins are all new, but thanks for the heads up. I'll check twice before running anything.
    Last edited by sanlyn; 21st Mar 2014 at 08:31.
    Quote Quote  
  13. Member
    Join Date
    Jul 2009
    Location
    Spain
    Search Comp PM
    Originally Posted by themaster1 View Post
    I have used Jmac's suggestion , using pointresize solve the chroma issue it appears.
    The current rgb>yv12 scheme is bugged to say the least.

    pointresize(last.width, last.height*4).converttoyv12(matrix="Rec601",inter laced=true).pointresize(last.width, last.height/1)
    Have you got a link to jmac's suggestion?
    If I understand it correctly (and I might be wrong), that bit of code will lose one of the fields.
    Unless perhaps it works to correct a bug (two wrongs making a right?).
    Quote Quote  
  14. here you go https://forum.videohelp.com/threads/340963-Best-quality-and-speed-video-denoisers-2011?...=1#post2127826

    It's a discussion you should be familiar with

    Doubling the height might be enough though, i have quadrupled it for some reason (probably because the result wasn't good, but now that i'm trying again it seems fine with a factor 2)
    Last edited by themaster1; 26th Jan 2012 at 15:34.
    Quote Quote  
  15. Member
    Join Date
    Jul 2009
    Location
    Spain
    Search Comp PM
    Originally Posted by themaster1 View Post
    Ah, I didn't realise it was that one, as that was in a different context and converting between YUY2 and YV12.
    Also, jmac698's original code was wrong as I pointed out in the post following.

    Doubling the height might be enough though, i have quadrupled it for some reason (probably because the result wasn't good, but now that i'm trying again it seems fine with a factor 2)
    To my mind, it shouldn't work either with quadrupling or with doubling, so I'm not sure why it does, unless it's compensating for some anomaly in NeatVideo/VirtualDub.
    Quote Quote  
  16. Member 2Bdecided's Avatar
    Join Date
    Nov 2007
    Location
    United Kingdom
    Search Comp PM
    Originally Posted by themaster1 View Post
    pointresize(last.width, last.height*4).converttoyv12(matrix="Rec601",inter laced=true).pointresize(last.width, last.height/1)
    I'm scratching my head what this will do to the interlaced chroma. If it is right, I wonder if it's two wrongs making a right, or just my total incomprehension!

    The full script:

    MPEG2Source("F:\00-videos_clips\MyVideo.d2v", cpu=0,info=3)

    ## DEBLOCKING INTERLACED SOURCE :
    par=getparity()
    SeparateFields().PointResize(width,height)
    Deblock_QED(24,28,uv=3)
    AssumeFrameBased()
    SeparateFields()
    Merge(SelectEven(),SelectOdd())
    par ? AssumeTFF() : AssumeBFF()
    Weave()
    That's weird. Wouldn't you just pointresize back, and weave? Rather than the last five lines above? Maybe it's another chroma issue I haven't thought through.

    Cheers,
    David.
    Quote Quote  
  17. The part regarding interlaced content deblocking was given to me by none other than Didée, that would be insulting toward him don't you think?
    Anyways, the problem is still here if i don't use the deblocking part so it's like i said a rgb->yv12 issue... most likely ...or a bug in neatvideo.

    I'm not certain but i believe neat video's internal colorspace by default is I420 (YVU)
    Quote Quote  
  18. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    Originally Posted by themaster1 View Post
    I'm not certain but i believe neat video's internal colorspace by default is I420 (YVU)
    That's what their user guide says. In the versions prior to v3 the manual gave fairly detailed info on that. You were allowed to choose a preferred color space, but YCrCb was the default. From the NeatVideo v1.1 manual, page 19 (also in later versions prior to v3):

    We recommend using the YCrCb color space to process all color video sequences
    in Neat Video. Normally, the YCrCb (default) working color space is best for
    color video sequences, the YCrCb Symmetric color space for grayscale (halftone)
    sequences represented in an RGB format. The RGB color space may be useful for
    special purposes, for example, to filter only one color channel (R, G or B) of
    the video sequence.
    A similar list of choices (but not quite the same) is in version 3, but the manual says little about the choices (page 10, version 3 manual):

    Neat Video converts RGB input data to the YCrCb working color space
    In v3 the "YCrCb Symmetric" choice doesn't appear:

    Image
    [Attachment 10757 - Click to enlarge]


    In v3 the color space is set to "Normal" by default.
    Last edited by sanlyn; 21st Mar 2014 at 08:31.
    Quote Quote  
  19. That's (highly likely) because the V3 now use The GPU (or CPU+GPU) and Nvidia cards (GT 220 for me) process the video in I420 at least that's what my card does on mediacoder with H264 (Cuda). If i have the misfortune to leave my script as is ( YV12) i'll get an error
    Quote Quote  
  20. Member 2Bdecided's Avatar
    Join Date
    Nov 2007
    Location
    United Kingdom
    Search Comp PM
    Originally Posted by themaster1 View Post
    The part regarding interlaced content deblocking was given to me by none other than Didée, that would be insulting toward him don't you think?
    Insulting, no, I just wanted to understand it. I've now unpicked it, and what it does it to keep the effect of QED deblock on the "invented" (duplicated) lines and the original lines, rather than throwing the effect on half of them away completely when re-interlacing (which a simple pointresize before weave would). I guess this must be beneficial. There are code fragments from Didée that I still don't fully understand the magic of, despite using them for nearly five years myself!

    btw, on paper, that 4*upscaled interlaced chroma resizing you're doing will only keep the chroma from one field - but whether that's what happens in practice, I don't know - point resize surprises me sometimes!

    Oh, I see Gavino said the same thing, and you said you're using 2* now. Let me try that...

    That's even worse - if all functions work as they should on paper, you're keeping 3/4 of the chroma from one input field, 1/4 of the chroma from the other input field, adding those two together, and then splitting that merged result over both of the output fields.


    I can't remember where this thread started now, so apologies if I'm talking nonsense - but AVIsynth converttoyv12(interlaced =true) works correctly in the vertical plane - no need to bodge it. The trick to get YUY2 into it losslessly by doubling the height isn't something that necessarily helps with NeatVideo (or other denoisers) because you're changing the noise pattern, and potentially doubling up individual bad pixels.

    I've never understood the obsession with YUY2 anyway. Unless you're processing in SD but upscaling to HD, all possible SD outputs have 4:2:0 chroma. Might as well work in that throughout and save yourself a headache. Unless there's some very specific weirdness that can only be removed in YUV2 but would still be visible in YV12.

    Cheers,
    David.
    Quote Quote  
  21. Any progress on this neatvideo issue ? Where are the avs scripts gurus when you need them ?
    Quote Quote  



Similar Threads