VideoHelp Forum
+ Reply to Thread
Page 4 of 4
FirstFirst ... 2 3 4
Results 91 to 99 of 99
Thread
  1. Originally Posted by sophisticles View Post
    ….. What they are doing is applying AQ principles to selling a home, but just as in AQ for video encoding it's all a scam, smoke and mirrors.
    Every encoder has a strategy how and where to spend the bits which constitute a limited ressource (bitrate constraints, file size constraints). Same with your house. If you have a limited quantity of paint you will have to think where you want to put the paint rather than just disperse it equally in the house, right?
    I wouldn't generalize that AQ does more damage than good. It was just my experience for this particular clip (meridien), using the NVEncC (Pascal) AVC encoder.

    Where you guys fail is you will take the PSNR of the source, ………..
    The source is the reference against which the encodes are compared. Being the reference the source alone has no PSNR.

    B.t.w. I posted the VMAF result of my NVEncC AVC encode rather than x264. Don't know how much NVEnc AVC relates to DS' et al. x264 encoder. Probably little to nil.
    Last edited by Sharc; 25th Oct 2019 at 05:53.
    Quote Quote  
  2. Dinosaur Supervisor KarMa's Avatar
    Join Date
    Jul 2015
    Location
    US
    Search Comp PM
    Originally Posted by poisondeathray View Post
    I probably miscommunicated . That was a contrasting example - My point was color splotches are digital noise. That is unwanted signal . It should say "nobody would argue that that was "digital noise"" . Bad phrasing

    This sample has wanted grain. Added on purpose.

    But the exact same pattern on another video might be "noise"

    The difference is intent and wanted vs. unwanted (by the producer, not necessarily the viewers)
    In my research on this video I can't find where they literally say they added noise/grain. I see where they picked certain shots, transitions, added smoke, etc. But I can't see where they say they added a layer of noise over everything. I mean the noise pattern I see is what I see when denoising the chroma on my DSLR photos in Lightroom. Left being no denoising on the RAW, right having Chroma denoising along with some light sharpening done to it. There is still noise that gets left behind.

    Image
    [Attachment 50553 - Click to enlarge]
    Quote Quote  
  3. Originally Posted by KarMa View Post
    But I can't see where they say they added a layer of noise over everything. I mean the noise pattern I see is what I see when denoising the chroma on my DSLR photos in Lightroom. Left being no denoising on the RAW, right having Chroma denoising along with some light sharpening done to it. There is still noise that gets left behind.
    They don't tell you everything they do; they didn't tell you they graded it either. Neither do studios releasing BD's . Grain is often added to BD"s too (and DVD's)

    One reason it looks slightly abnormal is this video sample is downscaled to 1920x1080 . The grain is less coarse, and smaller than usual

    It would be a terrible camera , if it had that amount of "noise" on a daylight exposed shot. Go download some native F65 or Weapon camera footage . Even low light footage shot with those has less "noise"

    Compare that to the BMPCC night time shot . That is unmistakably digital sensor noise

    Regardless of what it is - the job of an encoder is to emulate the source characteristics as closely as possible . So if you were testing BMPCC, those ugly chroma noise, color splotches are supposed to be reproduced .
    Quote Quote  
  4. [QUOTE=poisondeathray;2564476]
    Originally Posted by KarMa View Post
    Regardless of what it is - the job of an encoder is to emulate the source characteristics as closely as possible . So if you were testing BMPCC, those ugly chroma noise, color splotches are supposed to be reproduced .
    Absolutely. But here comes the point when 'uneducated' people may subjectively prefer the inferior encoder which suppresses such noise/details, rather than reproducing the source as closely as possible.
    Quote Quote  
  5. Originally Posted by Sharc View Post
    Absolutely. But here comes the point when 'uneducated' people may subjectively prefer the inferior encoder which suppresses such noise/details, rather than reproducing the source as closely as possible.
    You don't see the irony in your statement do you? For years, PDR, and those like him, have claimed that x264 was the best encoder because it looked the best and they also repeatedly took the position that you couldn't trust objective metrics like PSNR and SSIM, only your eyes. I was the one that repeatedly stated it's not the job of the encoder to produce the best looking encode but rather to produce the most accurate reproduction within the bit rate constraints set forth.

    Now that it suits his test results, he has adopted my stance.

    Intellectual honesty is definitely not a strong point of x264 proponents.
    Quote Quote  
  6. Originally Posted by sophisticles View Post
    You don't see the irony in your statement do you? For years, PDR, and those like him, have claimed that x264 was the best encoder because it looked the best and they also repeatedly took the position that you couldn't trust objective metrics like PSNR and SSIM, only your eyes. I was the one that repeatedly stated it's not the job of the encoder to produce the best looking encode but rather to produce the most accurate reproduction within the bit rate constraints set forth.
    WTF?! That's exactly what I've been saying for >10 years. Go back and look at the old series of tests. That's why x264 was rated on top, not because of some score. It most accurately resembled the input, whatever it may be . If there was grain or fine details, it kept the grain or fine details.

    The point was that PSNR and SSIM have low correlation rates with human perception, so you cannot trust them. The coefficient of correlation is like 0.3-0.4 for PSNR. Only mildly positive


    Now that it suits his test results, he has adopted my stance.
    WTF? It's not "your" stance, it's still my stance. Nothing has changed. This is what I've been saying all along

    If source has "grain" (noise if you want to call it) , and encode drops it, then it's NOT similar
    If source has tire details and encode drops it, then it's NOT similar
    If source has facial details, and encode blurs them away, that's NOT similar
    If source does not have banding/blocking in the sky, and encode suddenly does, that's NOT similar

    By by the same token, if encode has other artifacts, such as broken ratty lines, that's not similar either. If there is keyframe popping introduced, that's not similar either

    If you consider everything overall, some encodes are clearly more similar than others

    Intellectual honesty is definitely not a strong point of x264 proponents.
    You are the one being dishonest here, by misrepresenting others

    I've always said the job of encoder was to reproduce the input as best as possible . Go back and check those tests >10 years ago. It was long before you even registered under the Sophi handle
    Quote Quote  
  7. Originally Posted by poisondeathray View Post
    I've always said the job of encoder was to reproduce the input as best as possible . Go back and check those tests >10 years ago. It was long before you even registered under the Sophi handle

    There you go Sophi - This was a MPEG2 test, but same idea

    2009
    https://forum.videohelp.com/threads/305028-2009-MPEG-2-Encoder-Test!#post1871881

    the job of an encoder is to faithfully represent the frames sent to it, unless you set it not to do so, that should be the job of pre-processing and filters.
    Which handle were you under back then ? So you've been agreeing with my view then or adopted my stance since you registered in 2014. Why aren't you in agreement now ? Can you really not see it on this test or the dozens of others ?


    So how is x264 "overrated" ? I hear some people making these claims with zero evidence to support it. And if the same people making those types of claims cannot even tell the difference between the details of a low bitrate Youtube video vs. a higher quality, higher bitrate unrestricted AVC encode... I'm not sure anyone should take them seriously anyways.

    x264 it has weaknesses, sure it does. And other encoders do too, but stronger weaknesses.

    I go out of my way to collect scenes where x264 does poorly. It's just that other AVC encoders tend to do worse in those same scenes.

    x264 is on top of almost every test in the past 10 years for AVC . Subjective, objective. PSNR, SSIM, VMAF . If it was somewhere consistently in the middle sure, I'd call it overrated too. How can something be "overrated" when it is on top, or near the top of every test ?
    Quote Quote  
  8. Dinosaur Supervisor KarMa's Avatar
    Join Date
    Jul 2015
    Location
    US
    Search Comp PM
    Originally Posted by poisondeathray View Post

    They don't tell you everything they do; they didn't tell you they graded it either. Neither do studios releasing BD's . Grain is often added to BD"s too (and DVD's)
    ...
    It would be a terrible camera , if it had that amount of "noise" on a daylight exposed shot. Go download some native F65 or Weapon camera footage . Even low light footage shot with those has less "noise"

    Compare that to the BMPCC night time shot . That is unmistakably digital sensor noise
    Color grading is usually expected for any kind of high level video work. Adding noise is not expected, but denoising of some kind would be more likely imo.

    I downloaded 4 sample videos from RED's website. Instead of Weapon footage (which they don't list), I downloaded Dragon footage as a 6k replacement. https://www.red.com/sample-r3d-files

    Exposure Time: 5214
    ISO: 500
    https://files.videohelp.com/u/244047/A001_C018_1011KZ_0010.png

    Exposure Time: 4184
    ISO: 800
    https://files.videohelp.com/u/244047/A016_C001_02073O_0010.png

    Exposure Time: 1e+04
    ISO: 1000
    https://files.videohelp.com/u/244047/Q008_C037_11224I_0010.png

    Exposure Time: 2000
    ISO: 400
    https://files.videohelp.com/u/244047/J502_C086_01255A_0010.png

    You keep saying that the scene (Meridian) is in broad daylight but, it's a native 60fps so your exposure is going to max out at 1/120s maybe even faster. It has a very good depth of field (small aperture) so the light gathering is going to be hindered, and the sensor was probably hot in what looks to be California (ambient, running the camera a lot in failed takes, 60fps). Then there's the part where these RED RAW files are not 1:1 copies of the sensor data but instead lossy compressed sensor data, meaning there's probably a lot of denoising in my sample PNGs.


    Originally Posted by poisondeathray View Post
    Regardless of what it is - the job of an encoder is to emulate the source characteristics as closely as possible .
    Agreed, it is the job of the encoder to retain as much of the source as possible. My only gripe is the notion that most or all of the source of that "grain" is in digital post. And so I apparently can't call it noise but instead "wanted signal". Especially when this isn't based on input from the film maker.
    Quote Quote  
  9. Originally Posted by KarMa View Post
    Color grading is usually expected for any kind of high level video work. Adding noise is not expected, but denoising of some kind would be more likely imo.
    Adding grain occurs almost as frequently in post processing of high end digital acquisition as color grading - especially with today's digital acquisition (too "clean"), and especially with visual effects shots and compositing. Often scenes are degrained, processed, then regrained to match shots. People go to extremes to emulate that old grainy film look (for many types of releases). A lot of work goes into grain emulation technology, and grain plates


    https://www.creativeplanetnetwork.com/news/364113

    The team also incorporated a subtle, fine film grain, helping offset any “video” feel from digital motion picture camera HDR acquisition.
    My personal objection is that the "fine" film grain does not have the characteristics of commonly used film stocks. Usually the grain is coarser. When you add the downscale, I can see how people would consider it digital sensor "noise" . It's definitely not sensor noise (or at least that amount is not natural)
    Quote Quote  



Similar Threads

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