VideoHelp Forum
+ Reply to Thread
Page 2 of 2
FirstFirst 1 2
Results 31 to 43 of 43
Thread
  1. Member
    Join Date
    Jul 2007
    Location
    United States
    Search Comp PM
    @Livingonvideo

    While you may not see it as such, proper video restoration is an art that can never be replaced by automation. Many of the posters have spent years and sometimes decades perfecting the "art" of video restoration. It takes the human eye and mind to recognize and understand that the red sky and blue grass isn't an error that needs to be corrected, but a choice by the videographer making an artistic statement.

    As someone who is "a programmer, inventor and experienced in computer vision", you know that a program is only as good as the person programming it. Garbage In, Garbage Out. Any automated process is limited by the programmer's understanding and idea of what is correct.

    As I think about it now, that's where the disconnect between those who have posted (and given sound advice) and you occurs. You're trying to objectively quantify something that is nearly 100% subjective.
    Quote Quote  
  2. Banned
    Join Date
    Dec 2010
    Location
    New York
    Search Comp PM
    Manono, remember the good old days when I first used to post here? This reminds me of it.
    Quote Quote  
  3. Member
    Join Date
    Sep 2012
    Location
    Australia
    Search Comp PM
    I've been struggling to find a way to comment on this thread. In the throws of reeling in horror and recalling all the 'automatic post-processing' on my DVDs in the form of edge-enhancement and frame-rate conversions I've come to this conclusion:

    If you insist on doing a half-arsed restoration job on peoples private videos, for god sake, give them the before AND after in case they want it re-done properly later.
    Quote Quote  
  4. Member
    Join Date
    Sep 2012
    Location
    Australia
    Search Comp PM
    Originally Posted by Livingonvideo
    I am planning to start a restoration service for VHS and camcorder tapes here in Sweden,
    By the way, I believe the point here is to take peoples home videos in VHS and other legacy formats and convert them to digital files. 'Artistic effects' and 'film grain' doesn't really enter the discussion. Presumably people would want to discard the old tapes once they have a digital copy. Blindly filtering these videos may make them more watchable but WILL ALSO DAMAGE THEM, making them resistant to any future attempts at restoration.
    Quote Quote  
  5. Banned
    Join Date
    Oct 2004
    Location
    Freedonia
    Search Comp PM
    Originally Posted by Livingonvideo View Post
    I work with computer vision. According to your philosophy, this science is not possible either is it? How could a computer play chess or recognize images?
    I work in IT for a living. My university degree is in Computer Science. I also used to play in chess tournaments in the USA although at a very low level. For the benefit of the very few of you who will understand it, I played at class E. That's LOW. It basically means that I can usually beat any random dude who knows the rules but has never studied the game at all and simply knows how to play but I probably can't beat people who study the game. So all that makes me well qualified to comment on the above.

    You do not really understand how computers "play chess" at all. In the early days of computing in the 1950s and through maybe the 1980s, there was an assumption that the only way possible for computers to beat humans was to develop AI (Artificial Intelligence) systems that would mimic thinking as humans do it. But the problem with that is that AI is REALLY REALLY DIFFICULT. It's so difficult that all these years later, AI is still a far away goal that essentially we're not even in the same ballpark with. It might literally be centuries away before it can even begin to approach human type thinking.

    In the 1990s CPUs began to get quite a bit faster and storage started to get bigger and drop in price. So people working on what I will call "the chess problem" (getting a computer program able to beat a human) realized that they didn't have to do AI at all - not at all! - to build a program to defeat humans. Basically all they had to do was build a computer with massive computational capability and build massive databases of moves. This is not as difficult as you might think. There have been books for decades on the endgames where for certain combinations of pieces and positions, the moves to force a win are known. These books are quite large. I have a set. So basically all you have to do is provide the computer with the essential strategy like don't lose pieces for nothing, give them as much information as possible for known opening sequences so they avoid well known bad moves and give them massive end game lookup tables. Basically all the computer is doing is it's just calculating which move is likely to have the best outcome and playing it and when it gets down to the end, it just looks up the final moves to make. It's cheating. It's like taking a test where one person is allowed to consult every book they want (the computer) and the other person can only use their memory (the human). It's an unfair competition because there are limits to human memory. But computers are now greater than humans because the people at IBM cheated and found an easy solution (massive databases and massive computations to predict 10-20 moves later and determine what is better) instead of the better but more difficult solution (AI). Some benefit was gained in terms of learning better how to deal with looking up and selecting from massive amounts of information, but that's about it. In the end, computers won by cheating. In fact, it's so much cheating that one AI researcher invented the game Arimaa which uses a chessboard and chess pieces but has complete different rules and piece placement. Arimaa was deliberately designed so that computer scientists couldn't cheat like they did at chess and just build massive processing systems with gigantic databases and win by brute force. To date, the best human players in Arimaa can defeat the best computer programs.

    This is old so I can't say it represents the state of things today, but in the previous decade a lot of classic cartoons were ruined by using automated digital noise reduction which made idiotic decisions.
    http://www.cartoonbrew.com/classic/droopy-on-dvd-uncut-and-dvnrd-3285.html
    Things were so bad that at the time Disney spent the money to have its films cleaned up frame by frame by hand rather than get cheap and lazy and have the software at the time do a terrible job. So computers are not always great at these automated tasks.
    Quote Quote  
  6. Guest34343
    Guest
    Chess computers don't cheat. They follow the rules and have no more information than is available to a human who cares to look for it. Sure, they can remember more of that information, they can calculate deeper, and they can calculate faster, but it is not cheating. Based on what you said I would crush you (sorry about that). Will you say it's cheating because I remember openings and endings better and can calculate deeper and faster? It's just a matter of degree. If I can calculate some position exactly then it would be stupid to ignore my calculations and try to play in AI general-principle style. Finally, even computers have to evaluate static positions reached at the end of a calculation; it is not possible to calculate to a win/loss except for positions with limited pieces on the board. The evaluation function is very AI-like, in the sense that you use the term.

    What is so special about Arimaa's rules that precludes using databases and calculation? Bobby Fischer's idea of random chess at least eliminates most opening knowledge.

    If you want to say computers are cheating because they are not human, that is something else entirely, and pretty much devoid of any interesting content.

    Still, your point is useful but perhaps it is better said this way: The problem of winning a chess game is very different from the problem of restoration. The former is completely objective and the latter is largely subjective. A chess game is unambiguously won, lost, or drawn. Judging a restoration involves esthetic judgments and goals that may differ for different people, etc.
    Last edited by Guest34343; 7th Jun 2014 at 15:21.
    Quote Quote  
  7. Originally Posted by neuron2 View Post
    Still, your point is useful but perhaps it is better said this way: ....
    ....chess computers don't sue you for cheating. People will for crap workmanship, or false promise.

    Just something to consider.
    Quote Quote  
  8. Banned
    Join Date
    Nov 2005
    Location
    United States
    Search Comp PM
    since chess programs were brought up and one of my college majors was comp sci and i am a bit of a chess aficionado (i play every day and in fact it's the one computer game that i play most often), allow me to share this tidbit: you don't need AI or even a data base or library of chess openings, closings, middle game, positions or anything like that to code a chess game that can beat most players, all you need is simple math.

    what you do is assign each piece a value based on it's position, position relative to other pieces and the type of piece it is and you calculate a strength value for computer side and the players side. you then calculate the strength value again for all possible moves and simply choose the move that results in the highest value.

    this algorithm is very simplistic and was used by a stand alone electronic chess game that Radio Shack used to sell. that electronic game was strong enough that it was very difficult to beat, in fact my brother and i didn't beat it until he realized the algorithm it was using and that it would follow a set pattern based on the moves we made (it was calculating values and the same moves would result in the same values every time) so he found a way to "game" the algorithm and basically the match would result in the exact same moves ever time.

    this type pf brute force approach is good enough to beat most players.
    Quote Quote  
  9. Banned
    Join Date
    Dec 2010
    Location
    New York
    Search Comp PM
    Speaking of old tapes, I have a question regarding retail commercial DVDs that were obvious poor transfers of their original masters (pixelation, for example). How can I go about contacting the company and asking them if they would sell me a digital uncompressed copy of specific episodes? (this would be for home use only of course, and any editing/clean-up performed would be for me). If I give them the explanation, do you think such an arrangement can be made between a regular consumer and, say, Sony Pictures for instance?
    Quote Quote  
  10. Member
    Join Date
    Aug 2002
    Location
    Sweden
    Search PM
    There already are companies here in Sweden that can do this kind of video transfers so in what way would your service be better than the rest?

    If I wanted this kind of service I would like the option to not use your atuomatic postprocessing scripts. I beleive I can get better results with manual processing. Or at least offer both versions, the orignal and processed version. See answer from ndjamena above why i would want this.
    Ronny
    Quote Quote  
  11. Member PuzZLeR's Avatar
    Join Date
    Oct 2006
    Location
    Toronto Canada
    Search Comp PM
    Been away for a bit, but tune in sooner better than later.

    @Livingonvideo (if still around, or for anybody landing here on a search):

    Contrary to popular belief, an automatic post-processing script/technique is indeed arguably possible for VHS captures, once you sort out, and reduce, the chaos of different scenarios in the format going to digital.

    The real key is the VCR. If you have several tapes, and several capture VCRs, this branches it out. As one knows, some tapes play better, and even different, in one VCR over another, so results will be different in post-processing.

    For example, each VCR's, and settings', output will need different noise removal levels, chroma filters, motion algorithms and some may need dot crawl removal techinques (if an RCA/composite unit).

    Having said that, you can design a benchmark method for each VCR (as I have with my 4 capture units) - whichever one wins the job after initial testing on a given tape.

    If you are using only one VCR, this should be easier, however, a true professional will have several into the mix to diversify and optimize results.

    Of course, the investment in the learning curve to get your workflow tightened up and optimally efficient demands lots of pre-testing, and each filtering component is a minor science on its own. And there are other tasks necessary for film sources, deinterlacing, editing techniques, audio, etc, that need application in tandem with the regular processing, however such VCR-dependent workflows would indeed be a sound objective for someone like you seeking, both, a "formula", and a value added feature to your services.
    Last edited by PuzZLeR; 14th Jun 2014 at 12:10.
    I hate VHS. I always did.
    Quote Quote  
  12. Originally Posted by PuzZLeR View Post
    The real key is the VCR.
    Only partially, in my opinion. I have only the single VCR, but how I work with my material depends mostly on the material itself. I work with movies taped 20 years ago and those sources can be all over the place and what I do is entirely dependent on those sources. I have some things in common I do to most of them, but with different settings and strengths. I have a number of template scripts I use as a basis for what I'll be doing, but each is edited depending on what I think the source needs. Which is why I don't think, as you do, that "an automatic post-processing script/technique is indeed arguably possible for VHS captures".

    For the OP who's planning on a business, the VCRs used originally to create the tapes, and the tapes themselves (and maybe the capturing VCRs he'll use), will be of tremendous variety, which makes the auto-PP even less plausible. In my opinion.
    Quote Quote  
  13. Member PuzZLeR's Avatar
    Join Date
    Oct 2006
    Location
    Toronto Canada
    Search Comp PM
    I will partially agree.

    I believe, with the VCR specifics: dot crawl removal, chroma adjustments, noise removal - these can be automated into one script PER VCR used. Of course, if using several VCRs, you'd need one auto-PP script for each, but one such auto-PP script is indeed possible for each VCR IMO. At least arguably, and at least as a reasonable "average" in settings if one wishes to save time.

    And yes, after this initial layer of PP, it's not so automated when considering the content itself. I believe you're referring to film/interlaced/etc, crops, edits, A/R, etc. Yes, I agree here and have acknowledged it.

    So yes, auto-PP is possible in an initial layer (the VCR itself), and no, auto-PP is not possible when finalizing after (the actual content on the tape, independent of tape/VCR).

    At least that's how I see it.

    Actually, this is easier done than explained...

    (At any rate, this is more reason to work with lossless formats, or at least with DV, to make such adjustments with minimal penalty to quality.)
    I hate VHS. I always did.
    Quote Quote  
Visit our sponsor! Try DVDFab and backup Blu-rays!