Notices
Results 1 to 13 of 13
Like Tree1Likes
  • 1 Post By river_rat

Thread: Is data supercompression algorithm principally possible?

  1. #1 Is data supercompression algorithm principally possible? 
    Forum Professor
    Join Date
    Apr 2009
    Posts
    1,136
    What says modern theory about possibilty to compress video data (such as movies) father and father?
    Could we someday compress video thousands of times almost without looses? Where is the limit for compression?
    And what speed of processors would be required to decompess supercompressed data?


    Antislavery
    Reply With Quote  
     

  2.  
     

  3. #2  
    ***** Participant Write4U's Avatar
    Join Date
    Apr 2012
    Posts
    1,245
    Quote Originally Posted by Stanley514 View Post
    What says modern theory about possibilty to compress video data (such as movies) father and father?
    Could we someday compress video thousands of times almost without looses? Where is the limit for compression?
    And what speed of processors would be required to decompess supercompressed data?
    Comes to mind an approach that uses fractals.


    Reply With Quote  
     

  4. #3  
    Forum Professor
    Join Date
    Apr 2009
    Posts
    1,136
    Comes to mind an approach that uses fractals
    As I know this approach could be hardly called almost lossless.
    It could modify original image quite a lot and unpredictably?..
    Antislavery
    Reply With Quote  
     

  5. #4  
    Forum Professor river_rat's Avatar
    Join Date
    Jun 2006
    Location
    South Africa
    Posts
    1,510
    Any lossless compression function would have to be one-to-one (else it is not lossless). So suppose I had a magical compression function that could take n-bit strings and compress them to n-1 bit strings, this would result in a one-to-one mapping from a space of cardinality n to one of cardinality n-1, a contradiction.

    So there must exist incompressible n-bit strings.
    As is often the case with technical subjects we are presented with an unfortunate choice: an explanation that is accurate but incomprehensible, or comprehensible but wrong.
    Reply With Quote  
     

  6. #5  
    Forum Freshman
    Join Date
    Oct 2012
    Location
    Provence (South east of France)
    Posts
    93
    So there must exist incompressible n-bit strings.
    IMHO, random n-bit strings cannot be compressed without loss. By random, I mean each bit has a probability of 0.5 to be 0 or 1. But video images are not random, and video compression algorithms makes use of particular characteristics :
    • two successives frames are generally very similar, so you just have to record the differences between these two frames,
    • an image generally have areas containing pixel of the same color, so you can record this color and a definition of the area.

      The compression being based on repetitions, the maximum compression factor without loss depends on the degrees of repetition of the video:
    • a video showing only a black screen during one hour can be compressed with a rate very near 100%,
    • a "video" where each frame is totally independant of the previous and where each pixel is set randombly cannot be compressed.
    It is interresting to notice that, AFAIK , it is impossible to compress sampled sounds without loss, because each sample is independant of the previous. Of course, mp3 exists but it introduce some alteration of the sound.
    Reply With Quote  
     

  7. #6  
    Forum Professor river_rat's Avatar
    Join Date
    Jun 2006
    Location
    South Africa
    Posts
    1,510
    In general, questions of compressibility resolve into questions of Kolmogorov Complexity (kind of by definition). There are some interesting consequences in this field, for example a lot of things that seem random (digits of pi etc) are in fact highly compressible.

    I think the typical lossless sound compression ratio is 2:1 given the wav file.
    trfrm likes this.
    As is often the case with technical subjects we are presented with an unfortunate choice: an explanation that is accurate but incomprehensible, or comprehensible but wrong.
    Reply With Quote  
     

  8. #7  
    Forum Freshman
    Join Date
    Oct 2012
    Location
    Provence (South east of France)
    Posts
    93
    Thanks for correcting me, River_rat. What I wrote about sound was a nonsense. Sorry.
    Reply With Quote  
     

  9. #8  
    Ascended Member Ascended's Avatar
    Join Date
    Apr 2012
    Location
    Norfolk
    Posts
    3,411
    Quote Originally Posted by Stanley514 View Post
    Where is the limit for compression?
    There must be some limit of how much you could reduce complex sequences to make them simpler, but I would also think there is also a significant roll to be played by the computing power used to decompress it. The more computer power to available then the less actual data is needed to recreate an original signal. The computer and software should be able to fill in more and more of the data without it being needed in the orignal compression as they get progressively more powerful.
    Everything has its beauty, but not everyone sees it. - confucius
    Reply With Quote  
     

  10. #9  
    Suspended
    Join Date
    Jul 2012
    Posts
    613
    The general question about "data" has been answered by River Rat.

    You seem to be interested by video/sound so I will throw in some perspective: back in the day (20+ years ago), when computer were performing, you could store 1 hours of music on a floppy disk (1MB). Likewise your latest game engine can throw you 4000x2000 pixels at 90FPS. It wheight a few MB.

    But somehow, some lunies decided to SAMPLE sound instead of generating it. Any sampling is already a loss-full procedure. Compressed video regularly make my eyes bleed, but somehow, I should be content.

    Nowadays computers are parrots that play records (and badly). Lame.
    Reply With Quote  
     

  11. #10  
    ***** Participant Write4U's Avatar
    Join Date
    Apr 2012
    Posts
    1,245
    Reply With Quote  
     

  12. #11  
    Forum Junior JoshuaL's Avatar
    Join Date
    Mar 2011
    Location
    Seattle
    Posts
    268
    Boing3000, that's an interesting idea to generate from scratch rather than to play back. I imagine this would require a hefty amount of data to already exist on the "decode" side of the equation. That is to say, if instead of recording yourself talking and then me playing it back, you could send me just the text and my computer would read it back to me IN YOUR VOICE. But then presumably my computer would hold all the data required to imitate your voice, which might be a fair amount data. But I suppose you could find some sweet spot between how much data you are sending and how much is already stored on my end.... it's not compression in the usual sense, more like a word in a dictionary, where you could say a word is a "compression" of a more complex idea.
    "The only appropriate attitude for man to have about the big questions is... doubt. Doubt is humble, and that's what man needs to be, considering that human history is just a litany of getting sh*t dead wrong."

    Take two of these and call me in the morning
    .
    Reply With Quote  
     

  13. #12  
    Forum Bachelors Degree Kerling's Avatar
    Join Date
    Jul 2012
    Location
    Copenhagen
    Posts
    440
    Emancipation governs any further father compression.
    In the information age ignorance is a choice.
    Reply With Quote  
     

  14. #13  
    Suspended
    Join Date
    Jul 2012
    Posts
    613
    Quote Originally Posted by JoshuaL View Post
    But then presumably my computer would hold all the data required to imitate your voice, which might be a fair amount data.
    Yep voices emulation may need a lots of parameters, a guitar also. But has a general rule, you can be sure that is is always incredibly lower in bandwith than pure sampling. If you record 3 seconds of "Lalala" in one note, is weight 44100 * 3 * 16 bits (of uncompressed data) which is 264K of frozen/dead voice.

    Voice pitch corrector are common place nowadays, an in fact do a mathematical approximation of the voice to modify it. I would be surprise if a "I have a cold" plugin did not exist. Once and for all.

    I single track of loss-full MP3 is already 3MB (choose your loss at will). The java player on this site is probably less than that. but it is quite buggy. I suggest download the .mod and play them with VLC or Winamp. Both those "player" can play nearly every fromat at hand, from compressed to generated. They weight less than 50MB (skin and graphic visualizer for WinAMP) INCLUDED.

    I suggest trying this track which is has a good analogic "feeling" (25 year old .

    As there is music info, there could be light visualization info/rules. Once Html5 is enough of a standard, you'll see this old technology finally beat the crap of the "recording" industry. Things like Rome. Install Chrome hold your mouse and fasten your seat belt
    Reply With Quote  
     

Similar Threads

  1. Replies: 1
    Last Post: October 3rd, 2012, 12:21 PM
  2. Division algorithm
    By grautskaahl in forum Mathematics
    Replies: 2
    Last Post: May 31st, 2011, 12:54 PM
  3. need help on algorithm
    By itsme in forum Computer Science
    Replies: 0
    Last Post: May 23rd, 2010, 09:53 PM
  4. algorithm help please
    By CSlocote in forum Mathematics
    Replies: 1
    Last Post: November 19th, 2009, 07:30 PM
  5. Please hlep me in (Data preparation for data mining )
    By nadia in forum Computer Science
    Replies: 3
    Last Post: December 22nd, 2008, 01:56 PM
Bookmarks
Bookmarks
Posting Permissions
  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •