Alright, as the title might suggest, I am curious as to how these video compressions work. ASF (Advanced Streaming Format) I know is a streaming format made by microsoft, but other than that, I have no Idea as to how it works. Is this a frame by frame compression? RM (Real Media) I know is a compression that works by only storing aspects of a frame that have changed from the previous frame. However, I would like to know more about how this works. How can it tell what has changed from last frame? If the pixel is not exactly the same in the next frame, does it approximate? How does it say that a pixel should be the same as a previous frame without taking up as much space as a new pixel would? And why does it get pixely when there''s a lot of motion? This is the compression I''d like to know the most about. DivX (I have know idea why it is called DivX) I also know is a frame by frame compression based on MPEG 4, but how does it compress the video more than MPEG 4 does, and I still don''t get how MPEG 4 compression works either. Thanks in advance . I really hope someone knows how some of these compressions work, especially RM. This is essential to one of my projects. All info will be greatly appreciated! Thanks again

I think asf is MS''s version of MPEG-something

Ok, but do you know what compresses the mpeg more? Thanks for the response though!

All of them, ASF, DivX and RM are based on saving only the changed parts from the last frame. They also use DCT (same as JPEG is using) to make the data smaller. They discard information that human eye most likely wouldn''t notice - for example colour information is compressed very heavily.

But as for how RM exactly works, I bet you won''t ever find any information. It''s a commercial format made by a single company. Specs are probably classified.

Hmm, interesting. I always thought DivX was frame by frame! Shows how much I know. Do you know how the parts of images from the last frame are stored efficiently? Since DivX uses DCT, does it just check for repititions in the 8 x 8 blocks? Hmm... There is so much to learn. Arigatou

Alright, since DivX is the only non-corporation owned compression, how does DivX work, and how does the MPEG 4 compression, from which it is based, work. Also, how does the DCT compression work? I could never understand it from the explinations they gave in some docs from wotsit.org...

Thanks

If you really want to learn that stuff, I''d suggest not to take a too big bite. Learn JPEG compression first, it''s quite clever stuff too and you''ll need all the same knowledge in understanding MPEG compression.

DCT (discrete cosine transform) tries to find many cosine functions that sum up (interfere) to represent the image. There are libraries to find the cosine functions, for example
http://www.fftw.org/
Apparently DCT is calculated for 8x8 boxes.

There''s also a new compression method called wavelets (actually it was invented in the ''50s but it hasn''t been used until recently). Wavelets are a totally different method for lossy compressing and it''s much better approach than DCT. Only problem is that encoding/decoding wavelets needs a very fast processor. Eg. JPEG2000 is going to use wavelets for compression. Eventually, as computers get faster, wavelets will be used for video and audio compression too.

Thanks a lot. I''ll go check out the site and be on my way to understanding the two compressions. I''ll let you know if I have any more problems!

- IOFILE