VideoHelp Forum
+ Reply to Thread
Results 1 to 12 of 12
Thread
  1. Member
    Join Date
    Nov 2007
    Location
    I'm here
    Search PM
    Hi.

    Here is a sample of a video I have to make look reasonably nice. That means a lot less noise, while keeping the subject looking real, not too plastic. There's not much detail anyway, not much to lose, but I cannot get a satisfactory result. I tried TemporalDegrain, but at this source I find it will get to deformed when cranked up and lower settings will not improve much.

    Would anyone like to give a try?

    The original is huge, 50p fps at 38.3 Mbps, but thatīs what I have. Can it be made into 25p without simply throwing out every second frame, but getting some motion out of them?
    Image Attached Files
    Quote Quote  
  2. Originally Posted by Srivas View Post
    The original is huge, 50p fps at 38.3 Mbps, but thatīs what I have. Can it be made into 25p without simply throwing out every second frame, but getting some motion out of them?
    I have absolutely no idea what you mean. If you have 50p and you want to make it 25p (which sounds like a bad idea), by definition you will have exactly half the number of frames when you are finished with that conversion. To get to half the number of frames, you throw out every other frame. You don't get to magically maintain some of the original temporal information once you have half the number of frames. That is like saying, "throw away half the pixels in the image, but make it look as sharp as the high-res version."

    Thus, your statement of wanting to make that conversion without throwing out frames makes no sense.

    As for the video, it has obviously taken a very strange journey. I've never seen encoding artifacts like those, but they sure look like something that result from someone trying to get the video down to a really small size. It sure didn't happen when it was encoded to that super-high 38 mbps bitrate. Thus, it appears that the video has been encoded multiple times, with one of those encodes being done at a very low bitrate and with a very unusual encoder.

    Given the size of those noise blocks, I don't hold out much hope that you'll be able to do much to improve it.
    Quote Quote  
  3. Member
    Join Date
    Nov 2007
    Location
    I'm here
    Search PM
    A for the first part, I was wondering, is there a way to analyse the video, so that instead of simply throwing out every second frame, it's possible to get some temporal information from them and add it to the previous or next frame, or to both, to make 25p smoother. But even if possible, that is really secondary here and keeping 50p is no problem.

    And the video itself was initially badly recorded Iphone video, and the sample is a seriously zoomed in part of the original, to get the subject to the center. It's encoded only once and with a big bitrate, so that the original pixels are mostly there. The original is bad enough. But I meant to smooth out the artifacts, make it less aggressive, something like Neat Video is doing, but in avisynth.
    Quote Quote  
  4. To reduce the frame rate you could use QTGMC's motion blur feature. On the other hand, if the whole video is like this clip, there's not much motion, so why bother?

    For noise reduction, try TemporalDegrain().
    Quote Quote  
  5. Member
    Join Date
    Nov 2007
    Location
    I'm here
    Search PM
    Yes, no need to touch the framerate.

    Actually, although I tried TD already and wasnīt satisfied, this time I managed to get pretty good result. I guess thatīs pretty ok for such a video.

    TemporalDegrain(degrain=3, ov=6, blksize=16)

    What do you think? Can I tweak TD more to get something better out of it, without washing it out completely? Can I add some sharpening without amplifying the blocks, like to the edges?
    Image Attached Files
    Quote Quote  
  6. Member
    Join Date
    May 2014
    Location
    Memphis TN, US
    Search PM
    I guess there would be hundreds of suggestions. But in trying to remove 100% of the noise, 90% of the video went with it. The original is mostly noise. For sharpeners, I don't see anything to sharpen -- can't create detail from nothing. I gave it a try, using a config for TemporalDegrain that jagabo often suggested to "improve" without destroying (TemporalDegrain(SAD1=400, SAD2=200, Sigma=16). You can make it "look" sharper by resizing down (there ain't enough bitrate in the original for 1080p) and fixing the levels and gamma. Your sample has illegal brights and a high black point -- that's the main reason it looks washed out.

    Otherwise I don't see a lot of hope without wiping out everything. The attached is encoded for BD @25fps interlaced. I guess it could be tweaked.
    Image Attached Files
    - My sister Ann's brother
    Quote Quote  
  7. Move your phone closer next time. The zoom-in may look fine on the small screen but you have basically tossed out all your resolution.
    On the other hand, it's not like anyone's going to watch this on a big screen anyway.
    Quote Quote  
  8. It's not very effective with this low motion video but here's an example of how to use QTGMC's motion blur.

    Code:
    LSmashVideoSource("sample.mp4") 
    ColorYUV(gamma_y=40, off_y=-24) # lower black level, lighten shadows
    BilinearResize(640,360) # close to the real resolution of the video
    SelectEven() # discard every other frame
    QTGMC( Preset="Fast", InputType=1, ShutterBlur=1, ShutterAngleSrc=30, ShutterAngleOut=360, FPSDivisor=1) # motion blur the frames that are left
    TemporalDegrain(SAD1=300, SAD2=225, sigma=12) # I usually don't like to use more than 100,75,4 but this video is very bad
    aWarpSharp(depth=5)
    nnedi3_rpow2(4, cshift="Spline36Resize", fwidth=1280, fheight=720) # 4x upscale to 1280x720
    Sharpen(0.5)
    All the noise probably makes QTGMC's motion blur less effective. It might make more sense to move it after TemporalDegrain.
    Image Attached Files
    Quote Quote  
  9. Member
    Join Date
    May 2014
    Location
    Memphis TN, US
    Search PM
    Nice work. Always wondered what those QTGMC settings would do. Wish it wasn't necessary to toss 50% of the video, though.
    - My sister Ann's brother
    Quote Quote  
  10. Originally Posted by LMotlow View Post
    Wish it wasn't necessary to toss 50% of the video, though.
    You mean the frame rate? It wasn't really necessary. I just did it as an example of using QTGMC to get motion blur (the little you can see here). Regarding the severe denoising, we might get better results from the original video.

    Originally Posted by Srivas View Post
    the video itself was initially badly recorded Iphone video, and the sample is a seriously zoomed in part of the original
    Quote Quote  
  11. Member
    Join Date
    May 2014
    Location
    Memphis TN, US
    Search PM
    Originally Posted by jagabo View Post
    Regarding the severe denoising, we might get better results from the original video.

    Originally Posted by Srivas View Post
    the video itself was initially badly recorded Iphone video, and the sample is a seriously zoomed in part of the original
    I'm with jagabo and most others on this nagging point: just as with trying to work with botched dubs from VHS tape-to-tape, you'd get better results from the poorly recorded original than from a reprocessed, zoomed-in, resized, re-encoded copy. Another example of low-quality iPhone digital video made worse and needing industrial-strength rescue. That sample is the kind of digital stuff that looks hardly better than bad VHS.
    Last edited by LMotlow; 5th Nov 2017 at 03:24.
    - My sister Ann's brother
    Quote Quote  
  12. Member
    Join Date
    Nov 2007
    Location
    I'm here
    Search PM
    You guys are awsome! Thank you for your feedback and valuable advice!
    Really I'm just doing one tape ministry and when I'm not shooting myself, people bring me all kinds of stuff that I try to fix up.

    I managed to get the original, just to see will it really be possible to get something better out from the copy I had in the first place. Here is the sample, you will be surprised to see how it's taken, but it is what it is. It's a 59.94fps Iphone 6 video and the sample is cut with avidemux, as it is.

    The one I gave in the first place was converted to 50fps (since there's a good quality part of the whole video already in 50fps) and encoded in some higher bitrate to then be joined with the rest and encoded to the final video.

    Do you think it will make a difference in this particular case? Personally, I just do color correction and keyframing in Vegas Pro, then serve to avisynth for filtering and Megui to convert x264. This file has places to be turned around, resized, etc. and then joined with another clip from a Canon XA10 camera, a good quality part of it. So I don't know how to make it all in one render, especially when the frame rates are also different. In the case of an intermediate file, what is a good format to go for? I mean, if it's already bad, is there any point to encode into uncompressed or Lagarith etc. Will there be a big difference if a high bitrate avc format will be used as intermediate?
    Image Attached Files
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!