VideoHelp Forum
+ Reply to Thread
Page 2 of 3
FirstFirst 1 2 3 LastLast
Results 31 to 60 of 75
Thread
  1. Member Bernix's Avatar
    Join Date
    Apr 2016
    Location
    Europe
    Search Comp PM
    Probably Residual Combining Removal can mangle with original lines, but it shouldn't I think. Don't know. And it is not deinterlacer but related to deinterlacing.

    Bernix
    Quote Quote  
  2. Originally Posted by pandy View Post
    i'm not aware of existence any deinterlacer that affecting/alter present signal lines to recover missing ones
    Bob(), QTGMC(), all blend deinterlacers, all single frame rate deinterlacers...
    Last edited by jagabo; 17th Feb 2018 at 08:28.
    Quote Quote  
  3. Member Bernix's Avatar
    Join Date
    Apr 2016
    Location
    Europe
    Search Comp PM
    Hi Jagobo,
    isn't it comp has 1,3,5,7 lines and made 1,2,3,4,5,6,7 when blended so 1,3,5,7 are still untouched?

    Bernix
    Quote Quote  
  4. Originally Posted by Bernix View Post
    isn't it comp has 1,3,5,7 lines and made 1,2,3,4,5,6,7 when blended so 1,3,5,7 are still untouched?
    No, blend deinterlacers blur the two fields together. 0+1, 1+2, 2+3, etc. So you get what looks like double exposures (and you lose half the temporal resolution if you use single frame rate mode).
    Last edited by jagabo; 17th Feb 2018 at 08:55.
    Quote Quote  
  5. Member Bernix's Avatar
    Join Date
    Apr 2016
    Location
    Europe
    Search Comp PM
    And why then they don't works like I said. Much easier and effective I think. O.k. it is bob probably what I described, but this is right way to do it. But in some situation, when capture from our TV and using bob I get 50fps but frames are duplicated, sometimes I got 50fps unique. So in first case using yadif without bob.

    Bernix
    Quote Quote  
  6. Originally Posted by Bernix View Post
    And why then they don't works like I said. Much easier and effective I think.
    It would leave comb artifacts and look terrible. One field would be untouched, the other would be a blend of two fields.

    Or maybe you're describing a bob deinterlace? Some preserve the original field, some don't.
    Quote Quote  
  7. Member Bernix's Avatar
    Join Date
    Apr 2016
    Location
    Europe
    Search Comp PM
    I disagree, you already said both fields comes from different times, if i interpolate lines 1st and 3rd 1(2)3 in this case i got 2 it will do in each frame and each line. Don't understand why one field should be untouched or blending fields. Just in each fields interpolate missing lines from field to get full frame. In first even in second odd. Depending on fields order. And this is lossless deinterlacing in meaning that 1st and 3rd lines remain untouched.
    Bernix
    Quote Quote  
  8. You're describing a bob deinterlace, not a blend deinterlace.
    Quote Quote  
  9. Member Bernix's Avatar
    Join Date
    Apr 2016
    Location
    Europe
    Search Comp PM
    And why then they don't works like I said. Much easier and effective I think. O.k. it is bob probably what I described, but this is right way to do it.
    It was in the post where I describe the method and then you tells me it is wrong, some untouched some weird and ugly. So therefore I don't understand what method do you mean.

    Edit: And as blending I understand blend lines in field, not fields together But it is sad, it really is used to blend fields....

    Bernix
    Last edited by Bernix; 17th Feb 2018 at 09:28. Reason: Edit
    Quote Quote  
  10. Video Restorer lordsmurf's Avatar
    Join Date
    Jun 2003
    Location
    dFAQ.us/lordsmurf
    Search Comp PM
    Originally Posted by pandy View Post
    Originally Posted by LMotlow View Post
    Originally Posted by pandy View Post
    Deinterlacing is lossless operation from video quality perspective.
    Not true, and not possible. The laws of physics don't support that statement. The second part of the above post is obvious.
    It is true and inline with math and physics - your claim is like blaming interpolator for not being able to extend signal bandwidth... (hope you understand why resampling for example 22kHz signal to 44kHz will not extend signal itself bandwidth). Deinterlacing is special form of signal resampling (interlacing is founded on principle where every second sample is removed and this violate Nyquist law) and perfect deinterlacer doesn't exist as quest for perfect reconstruction of signal before sampling is more like artistic educated guess than linear math.
    There is loss no matter what.
    Don't agree? Take up it with Mr. Faroujda himself.
    Want my help? Ask here! (not via PM!)
    FAQs: Best Blank DiscsBest TBCsBest VCRs for captureRestore VHS
    Quote Quote  
  11. Member Bernix's Avatar
    Join Date
    Apr 2016
    Location
    Europe
    Search Comp PM
    Bob is lossless, of course if you encode in low bitrate xvid or any compression except lossless, you can't get original anymore. But you have 50 half-frames (fields) per second, so create 25 full frames from them is against humanity. Don't forget what jagabo said, it is fields from 50 different times. That is all. Can be problem with 23,976 to bob (don't know) it and get some standard framerate, but in PAL bob is only correct way to deinterlace. Better to keep it untouched then blend fields....

    Sincerely Bernix
    Quote Quote  
  12. AviSynth's Bob() is lossy by default. You can make it lossless with Bob(0.0, 1.0) -- but only for RGB and YUY2 (not for YV12).
    Quote Quote  
  13. Originally Posted by lordsmurf View Post
    There is loss no matter what.
    Don't agree? Take up it with Mr. Faroujda himself.
    Loss of what? Non existing signal? You compare deinterlaced signal to? Progressive source?
    This thread looks to me like another audio-voodoo (only this time video voodoo) - fashionable to complain on lossy deinterlacing - some audiophiles reject oversampling concept - there is large church of NOS (Non OverSampling) DAC's and same fishy smell on deinterlacing - if deinterlacing is lossy then real videophiles should fill missing lines in fields by black (gray, white, whatever) - there will be no loss so perfect. Mr Faroudja sold his video business long time ago... seem new owner is also owner of Nova Perseus...
    Quote Quote  
  14. Pandy's original broad claim that "Deinterlacing is lossless" is obviously not true. His later refinement that it's lossless when the original fields are left intact is true if you also assume double frame rate methods (30i to 60p, 25i to 50p). But many deinterlacers do not double the frame rate and even those that do often don't leave the original fields intact. Samples of both have already been given in this thread. No doubt, he will probably now tell us that deinterlacers that don't preserve the original fields aren't really deinterlacers.

    In any case, that lossless aspect is irrelevant when the goal is to make interlaced video look "good" on progressive displays. I would much rather have a deinterlacer that doesn't perfectly preserve the original fields but looks good (QTGMC's default, for example) over one that preserves the original fields but looks like crap (duplicate field, for example).

    His other claim that you can't fully restore the vertical bandwidth of the image is true -- for deinterlacers that work with only one field at a time. All the better deinterlacers use both fields of the frame, even fields from other frames, and motion interpolation to fill in the missing fields. In the best case these can double the vertical bandwidth. In the worse case they can't. In most cases, it's somewhere in between.

    Pandy knows all of this, of course. He just want's to be pedantic and pigheaded.
    Quote Quote  
  15. Video Restorer lordsmurf's Avatar
    Join Date
    Jun 2003
    Location
    dFAQ.us/lordsmurf
    Search Comp PM
    Originally Posted by jagabo View Post
    Pandy's original broad claim that "Deinterlacing is lossless" is obviously not true. His later refinement that it's lossless when the original fields are left intact is true if you also assume double frame rate methods (30i to 60p, 25i to 50p). But many deinterlacers do not double the frame rate and even those that do often don't leave the original fields intact. Samples of both have already been given in this thread. No doubt, he will probably now tell us that deinterlacers that don't preserve the original fields aren't really deinterlacers.

    In any case, that lossless aspect is irrelevant when the goal is to make interlaced video look "good" on progressive displays. I would much rather have a deinterlacer that doesn't perfectly preserve the original fields but looks good (QTGMC's default, for example) over one that preserves the original fields but looks like crap (duplicate field, for example).

    His other claim that you can't fully restore the vertical bandwidth of the image is true -- for deinterlacers that work with only one field at a time. All the better deinterlacers use both fields of the frame, even fields from other frames, and motion interpolation to fill in the missing fields. In the best case these can double the vertical bandwidth. In the worse case they can't. In most cases, it's somewhere in between.

    Pandy knows all of this, of course. He just want's to be pedantic and pigheaded.
    I've honestly never considered "separation of fields" to be deinterlacing. For one thing, you're given a half-height image. Separation is simply a processing step, while true deinterlacing is converting it to a watchable progressive format and the correct frame size. And with it comes loss -- varying from substantial and awful, to something acceptable like QTGMC.

    There's also a huge misunderstanding about framerate doubling. When you actually deinterlace (meaning NOT merely separating field), it interpolates from neighboring fields/frames. So that "59.94p" files made from 59.94i/29.97fps is false. Interpolation creates data. QTGMC, for example, is an EDI at its core (where "i" means interpolation). By choosing to return to the 29.97fps rate, you're not really losing much/any temporal data. Sure, during really fast motions, maybe some interstitial motions, but nothing that detracts from the overall body of work. If you cycle through all the frames, a vast majority will merely be doubled data. This is comparable to what an HDTV does. You can keep the 59.94p if you insist, but the main drawback is that you now have a file that is 200% bigger, and invalid for most formats/players. Great for forensics, not so much to sit on the couch with a bowl of popcorn.

    Of course, you (jagabo) know all this. But others ... not so much.
    Want my help? Ask here! (not via PM!)
    FAQs: Best Blank DiscsBest TBCsBest VCRs for captureRestore VHS
    Quote Quote  
  16. Member Bernix's Avatar
    Join Date
    Apr 2016
    Location
    Europe
    Search Comp PM
    You need much more than 200% if you mean lossless. And if you mean compress you need less than 200% because it is well known that progressive compression is more effective. + double framerate (probably better to say (field to frame) because speed is same, the compression can be much more effective because smaller difference between frames. Much more effective P (reference frames) and ME at least.

    Edit: where you get lossless (not already compressed video) On tv hardly. From miniDV camera also hardly.

    Bernix
    Last edited by Bernix; 18th Feb 2018 at 03:45. Reason: Edit
    Quote Quote  
  17. Originally Posted by lordsmurf View Post
    There's also a huge misunderstanding about framerate doubling. When you actually deinterlace (meaning NOT merely separating field), it interpolates from neighboring fields/frames. So that "59.94p" files made from 59.94i/29.97fps is false. Interpolation creates data.
    Interlaced NTSC is 59.94 fields per second. I'm not sure I understand how de-interlacing to 59.94 frames per second would be any more false than de-interlacing to 29.97 frames per second.

    Originally Posted by lordsmurf View Post
    QTGMC, for example, is an EDI at its core (where "i" means interpolation). By choosing to return to the 29.97fps rate, you're not really losing much/any temporal data. Sure, during really fast motions, maybe some interstitial motions, but nothing that detracts from the overall body of work.
    I'd have to disagree with that one. Maybe more so for PAL than NTSC, given PAL is 50 fields per second vs 59.94 for NTSC.
    Single frame rate de-interlacing can look "jittery" compared to double frame rate deinterlacing, one reason being video doesn't have the same motion blur as film. QTGMC has settings to add motion blur, but 50/59.94fps still looks smoother.

    Originally Posted by lordsmurf View Post
    If you cycle through all the frames, a vast majority will merely be doubled data. This is comparable to what an HDTV does.
    If the source is truly interlaced, there won't be duplicate frames anywhere there's motion (double frame rate de-interlacing). There will be frames that are close to duplicates where there's zero motion, but that applies to single frame rate de-interlacing too. Duplicate frames take very little bitrate to encode anyway.

    Originally Posted by lordsmurf View Post
    You can keep the 59.94p if you insist, but the main drawback is that you now have a file that is 200% bigger, and invalid for most formats/players. Great for forensics, not so much to sit on the couch with a bowl of popcorn.
    Not true. I've run test encodes (PAL sources) with x264 at CRF18, and depending on the de-interlacer, the file size increase is somewhere between 5%-10% (from memory). It does depend on the de-interlacer though. In my tests, the resulting file sizes after de-interlacing with QTGMC at 50fps were usually lower than after deinterlacing with Yadif at 25fps.
    There's some examples attached to this post if you want to confirm that for yourself. There's a fairly obvious difference between 50fps and 25fps in respect to motion too, especially for Yadif, and there's an aliasing reduction at 50fps compared to 25fps.

    Even though I reduced the resolution to 640x480 from 720x576, the QTGMC 50fps version looks better than the source does when my TV is doing the de-interlacing. I'd be interested to learn how they compare with the source for johnmeyer using his TV's de-interlacing.

    Every player in our house (Bluray players and TVs with built-in media players) support Level 4.1, even if not always at the maximum bitrate, but they all play standard definition 50fps/59.94fps h264 without a problem. I think the same applies to 720p, given it's still Level 4.1 complaint at 50fps/59.94fps, although I generally use my PC for playback so I'd have to check that one. Most newer TVs, especially 4k models, would no doubt support higher levels.
    https://en.wikipedia.org/wiki/H.264/MPEG-4_AVC#Levels

    Personally I don't care if the original fields aren't retained as long as the final result looks better. I use QTGMC in progressive mode as a noise filter quite a bit, and when used that way it throws away ever second scanline and interpolates new ones, but for many sources I find it can be more effective than other noise removal plugins/scripts. Not perfect for sure, but the detail loss/noise removal ratio tends to be quite good, especially where there's motion.
    Last edited by hello_hello; 18th Feb 2018 at 05:49.
    Quote Quote  
  18. Member Bernix's Avatar
    Join Date
    Apr 2016
    Location
    Europe
    Search Comp PM
    Here I create some graphs how i understand lossless deinterlacing.
    Here is original progressive frame.
    Image
    [Attachment 44727 - Click to enlarge]

    Here is field from this progressive image
    Image
    [Attachment 44728 - Click to enlarge]

    And here is linear interpolation of missing lines
    Image
    [Attachment 44729 - Click to enlarge]

    It is not perfect, it is just linear interpolation. But how and why to demolish original lines from one field picture? You can get them from result back. And yes, linear interpolation isn't probably best, but even other non linear methods should keep original values, and don't change them.

    Of course you can refine missing lines from previous - next fields, where missing lines can lean on existing lines (they are also intact there). Question is how, but still lossless.

    Bernix
    Last edited by Bernix; 18th Feb 2018 at 08:45.
    Quote Quote  
  19. Originally Posted by hello_hello View Post
    I've run test encodes (PAL sources) with x264 at CRF18, and depending on the de-interlacer, the file size increase is somewhere between 5%-10% (from memory).
    The main reason the increase is so small is because x264 uses higher quantizers with higher frame rates. Ie, 60p video is encoded with lower quality (per frame) than 30p video.
    Quote Quote  
  20. Originally Posted by jagabo View Post
    Pandy knows all of this, of course. He just want's to be pedantic and pigheaded.
    lol - you can't resist to not go for argumentum ad personam... truly flattered with so much attention to my humble person not to problem.
    Of course you know what is purpose for interlacing (exchange spatial and temporal resolution matched with human perception system at bandwidth constraints). Cases mentioned by you are special cases and best way to avoid de-interlacing is not applying interlace first (i assume you will not recommend to deinterlace progressive movie).
    Btw - higher temporal resolution leading to more accurate motion coding i.e. higher coding gain...

    Sorry guys for touching videophiles nest... good luck with banality of claims 'deinterlacing is lossy' .
    Last edited by pandy; 18th Feb 2018 at 17:32.
    Quote Quote  
  21. This is a very tortuous thread that conflates many issues with many misstatements by just about everyone. I have personally vetted numerous endlessly lossless pathways in my workflows. It is a very exacting process and complicating it with a filter like deinterlacing is just foolishness. However, theoretically, one could losslessly deinterlace (assuming here the criteria for lossless is the ability to perfectly regenerate the source video from the deinterlaced video) if one used exact interpolation and preserved all the fields--but this looks like crap, so it is never done in practice, plus it would require a special deinterlacer devoid of any AA/dithering or smoothing. Just because Avisynth has been tuned to behave a certain way does not mean that is how all cases behave or could behave. IOW, what people really mean when they argue on VH about such and such being lossless is Avisynth's implementation is not lossless.
    Quote Quote  
  22. Originally Posted by SameSelf View Post
    This is a very tortuous thread that conflates many issues with many misstatements by just about everyone. I have personally vetted numerous endlessly lossless pathways in my workflows. It is a very exacting process and complicating it with a filter like deinterlacing is just foolishness.
    Do you watch your interlaced video on a progressive display without foolishly de-interlacing it on playback?

    Originally Posted by SameSelf View Post
    However, theoretically, one could losslessly deinterlace (assuming here the criteria for lossless is the ability to perfectly regenerate the source video from the deinterlaced video) if one used exact interpolation and preserved all the fields--but this looks like crap, so it is never done in practice, plus it would require a special deinterlacer devoid of any AA/dithering or smoothing.
    In practice, it's exactly what Yadif(mode=1) does.
    It's also exactly what QTGMC(Lossless=1) does.
    In practice QTGMC(Lossless=2) restores the original pixels before applying any sharpening or temporal smoothing.

    Originally Posted by SameSelf View Post
    Just because Avisynth has been tuned to behave a certain way does not mean that is how all cases behave or could behave. IOW, what people really mean when they argue on VH about such and such being lossless is Avisynth's implementation is not lossless.
    Bob() is the built in Avisynth de-interlacer, but nobody has recommending using it. Did you read the thread before deciding what everyone else means?
    Bob(0.0, 1.0) is lossless for RGB and YUY2 and preserves the Luma for YV12.
    Quote Quote  
  23. Originally Posted by johnmeyer View Post
    Deinterlacing ALWAYS If your video is interlaced, save it as interlaced and then use the deinterlacer in your TV when you watch it.
    You're suggesting the cheap IC-based, real-time deinterlacers in consumer electronics will produce a better output than a properly configured, offline, non-real-time deinterlacer.

    What absolute nonsense.
    Quote Quote  
  24. Originally Posted by EternalQuadrangle View Post
    Originally Posted by johnmeyer View Post
    Deinterlacing ALWAYS If your video is interlaced, save it as interlaced and then use the deinterlacer in your TV when you watch it.
    You should probably get your quotations correct.

    Originally Posted by johnmeyer View Post
    Deinterlacing ALWAYS degrades the video.
    .
    .

    My recommendation has always been this: do NOT deinterlace unless you absolutely have to. If your video is interlaced, save it as interlaced and then use the deinterlacer in your TV when you watch it.
    Originally Posted by EternalQuadrangle View Post
    You're suggesting the cheap IC-based, real-time deinterlacers in consumer electronics will produce a better output than a properly configured, offline, non-real-time deinterlacer.

    What absolute nonsense.
    Unless using AviSynth's QTGMC as a deinterlacer then, yes, a hardware deinterlacer could very well be as good or better than any other software deinterlacer. They're pretty good these days, and getting better all the time.

    Did you sign up here just to make that misinformed comment on a thread nearly a year old?
    Quote Quote  
  25. Originally Posted by EternalQuadrangle View Post
    Originally Posted by johnmeyer View Post
    Deinterlacing ALWAYS If your video is interlaced, save it as interlaced and then use the deinterlacer in your TV when you watch it.
    You're suggesting the cheap IC-based, real-time deinterlacers in consumer electronics will produce a better output than a properly configured, offline, non-real-time deinterlacer.

    What absolute nonsense.
    Who are you? You can't even quote me correctly, something that requires nothing more than a click of the mouse. You actually have to work at getting it wrong.

    And your comment neglects the fact that the hardware deinterlacer has direct control of the hardware itself. Also, it may be cheap, but that has nothing to do with whether it can get the job done well. It has been a long, long time since I've seen anyone complain about the quality of deinterlacing on a Sony, Samsung, or other quality TV. Those days are long gone.
    Quote Quote  
  26. aBigMeanie aedipuss's Avatar
    Join Date
    Oct 2005
    Location
    666th portal
    Search Comp PM
    those CHEAP ic "integrated circuits" are what power computers also, and they are the brains of all hardware. ignore the ignoramus, it was it's first post here and likely it's last.
    --
    "a lot of people are better dead" - prisoner KSC2-303
    Quote Quote  
  27. Perhaps a bit off topic, but once we are at it:
    Does anybody know how today's TVs or HW deinterlacers process one field phase-shifted "interlaced" video? Do they match the fields and hence restore the original progressive frames perfectly, or do they deinterlace it (producing unavoidable artefacts)?
    Last edited by Sharc; 26th Feb 2019 at 02:52.
    Quote Quote  
  28. Originally Posted by Sharc View Post
    Perhaps a bit off topic, but once we are at it:
    Does anybody know how today's TVs or HW deinterlacers process one field phase-shifted "interlaced" video? Do they match the fields and hence restore the original progressive frames perfectly, or do they deinterlace it (producing unavoidable artefacts)?
    The vast majority are 'flag readers' and will deinterlace. The better hardware chips in DVD/Blue-ray players and televisions are called 'cadence readers' and will reassemble the fields into progressive frames frames.

    Unfortunately, the situation is pretty grim for PAL people, since most DVDs are encoded as interlaced and tmost players read only that interlace flag to tell it what to do.
    Quote Quote  
  29. Originally Posted by manono View Post
    Originally Posted by Sharc View Post
    Perhaps a bit off topic, but once we are at it:
    Does anybody know how today's TVs or HW deinterlacers process one field phase-shifted "interlaced" video? Do they match the fields and hence restore the original progressive frames perfectly, or do they deinterlace it (producing unavoidable artefacts)?
    The vast majority are 'flag readers' and will deinterlace. The better hardware chips in DVD/Blue-ray players and televisions are called 'cadence readers' and will reassemble the fields into progressive frames frames.

    Unfortunately, the situation is pretty grim for PAL people, since most DVDs are encoded as interlaced and tmost players read only that interlace flag to tell it what to do.
    Ah yes, flag reading vs cadence detection, I remember now.
    Fortunately, I think, the situation is somewhat less critical for commercial DVDs as interlaced encoded DVDs are normally not phase-shifted but the fields are taken from the same original frame (no time shift between the 2 fields). I was surprised though to find Vegas - under certain circumstances - to produce field phase shifted video when converting progressive footage to interlaced video. As the playback device just deinterlaced it (flag reading, apparently) the result was sub-optimum with typical deinterlacing artefacts.
    Quote Quote  
  30. Originally Posted by Sharc View Post
    Perhaps a bit off topic, but once we are at it:
    Does anybody know how today's TVs or HW deinterlacers process one field phase-shifted "interlaced" video? Do they match the fields and hence restore the original progressive frames perfectly, or do they deinterlace it (producing unavoidable artefacts)?
    Most of modern reputable brands TV's are equipped with some form of motion detection so they should be able to deal with odd cases (but not always without issues). And TV's usually don't know flag (i mean by this that multimedia and TV reception functions are usually processed in different circuit than video thus video processor almost always perform blind deinterlacing). It is mandatory to have motion detection in all framerate converting TV's (and many of them are in fact asynchronous internally with fixed refresh rate when compared to incoming video). IMHO Better brands trying almost everything in approach toward good deinterlacing.
    (and they are quite good, can observe SD interlace on very old LG TV, and interlace they can be seen only in very fast motion areas with relatively low contrast like light gray on white - then some lines can be visible but on limited area - this LG TV is at least 6 years old and modern TV should be way better).
    Quote Quote  



Similar Threads

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