h a l f b a k e r yNot so much a thought experiment as a single neuron misfire.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
For scenes such as lectures, speaches, news casts etc, where frames don't change much, one can safely say that first few seconds of video contain all or most of the required frames for remaining portion of the video. Hence instead of replicating the frames, in subsequent portion of the video, simply
refer to earlier frames. This could be very useful in streaming videos.
First few ( say , 2) seconds of videos will contain actual frames. Later frames will instead simply give the index no. of the frame within first 2 seconds. This video of first 2 seconds will be used as a lookup table.
One scenario where this could be used will be videoconferencing between two people in their homes. Since they are in their homes, their background won't change at all, in most cases. The actual changes will contain mostly facial expressions, lip movement etc. which are repeatative. In such cases it might be possible to send realtime HD video over low bandwidth connection.
[edit]:
One major advantage will be : Once first look up tabled has been streamed/loaded, for later portion of the video, resolution is irrelevant; It could be 320 * 200 or 4000 * 3000; It will be equaly efficient.
Blue
http://www.imdb.com/title/tt0106438/
Just what it says in the title. [8th of 7, Jul 12 2011]
[link]
|
| |
//This could be very useful in streaming videos.// |
|
| |
Compression schemes for video are usually designed so that you can start playing a video from somewhere in the middle of the stream. |
|
| |
If your scheme refers back to the first 2 seconds of video then it's not possible to decompress the stream unless you've seen the first 2 seconds of it. |
|
| |
This is kind of what MPEG does - the bits of the frame which don't change are only updated infrequently. |
|
| |
//start playing a video from somewhere in the middle of the stream//
In that case, first two seconds of video will have to be downloaded anyway. |
|
| |
This already happens though, in all sorts of existing codecs, does it not? [edit] - ahh yes - what [hippo] said. |
|
| |
I think what hippo says is diferent than what is being proposed. What is suggested here, is to preserve first two (or whatever) seconds of video even after displaying it, which I believe curretnly does not happen. Currently, portion of video which is displayed is removed from the memory. |
|
| |
IN the new scheme, if video length is 2 hours, first few seonds of video will be preserved in memory ( probably high speed graphics memory)for 2 hours. |
|
| |
//For scenes such as lectures, speaches, news casts
etc, //
I think you'll find all of these have at least slightly
changing backgrounds, if that is not important, then
just send the first frame and play the audio. |
|
| |
//In that case, first two seconds of video will have to be downloaded anyway.// |
|
| |
Nope. Have you ever noticed that digital TV still works if you turn the telly on half way through a programme? |
|
| |
What I've noticed is that if I turn on or start a video over the internet (and sometimes the effect is slightly visible on some lower quality DVDs) that in 'static' scenes, you can get a wash of colour that is from some previous scene (or some previous point in the current scene) that is only modified/swept clean when some clear-contrast change passes over that portion of the picture - or, in other words, "Yes, I have noticed that Internet TV sometimes has artifacts that would suggest that some kind of static-area compression is in effect" perhaps not from the start, but more often a kind of periodical effect. |
|
| |
This would work very well for some movies .... |
|
| |
//An idea that is based on complete ignorance of the subject matter// Explain. |
|
| |
We think you just did ... |
|
| |
It'd work for colour palettes though I'm not sure if they do that these days. Cartoons and other animation also. |
|
| |
I agree there must be a more efficient way to compress video recorded from a stationary camera. |
|
| |
My idea would be to record the background as the 'look-up table' and then only have to compress and transmit the moving foreground. Actually, I'm pretty sure I read about exactly this type of video compression in a magazine a few years ago. |
|
| |
I didnt see how this was different from regular compression at first but now I do. [+] |
|
| |
//from regular compression but now i do. |
|
| |
mp4? thats exactly how it works. The algorithm of course is more detailed. (I was on the mpeg implementation team) |
|
| |
//This is kind of what MPEG does - the bits of the frame which don't change are only updated infrequently.// |
|
| |
This is where this new idea has a major plus.
In this case, even if frames do keep changing, even at high speeds for hours together, it is still fine, as long as frames are available in the video lookup table. I think mpeg will fail badly in this case. |
|
| |
This would work well for most soaps. |
|
| |