Originally posted by michaelab
Can't remember what MPEG1 is but MPEG2 is used for video compression and is the format in which all DVDs are stored.
As with all MPEG standards, MPEG-2 is actually a collection of standards for video and audio, not just video.
MPEG-1 was the prior standard, again for video and audio.
MPEG-1 was not designed with broadcast media in mind so it is not very error tolerant. It also does not cope as well with interlaced pictures as MPEG-2 (in fact, dealing with interlaced pictures was strapped on as an appendix in MPEG-1).
Originally posted by michaelab
MPEG3 (usually known as MP3) is a compression format for music. MPEG4 is a newer (and better?) compression format for video.
There is no such thing as MPEG-3.
MP3 is in fact, Layer 3 of the MPEG-1 Audio standard.
MPEG-2 Audio is backwardly compatible with MPEG-1; therefore MP3 can be used to refer to MPEG-2 Audio, but only those elements which are equivalent to MPEG-1 Audio Layer 3. MPEG-2 also provides AAC (advanced audio coding) which significantly improves on MP3 compression.
Originally posted by michaelab
MPEG4 is a newer (and better?) compression format for video.
Yes, but again, not just video. MPEG-4 was developed with computer applications and mobile broadcasting in mind; it therefore deals not just with video and audio, but also with graphics, textures, speech and so on. To do this, it differs radically in how it compresses images and audio; therefore achieving vast improvements over MPEG-1 and -2 (it takes an "object-oriented approach" to it all).
Originally posted by michaelab
In this way it can achieve incredible compression ratios (like 1:10 or even 1:20).
30:1 is typical for MPEG-1. MPEG-2 is even higher.
Originally posted by michaelab
I don't know exactly how MPEG/JPEG image compression works but it starts by dividing an image into square blocks - the bigger the blocks, the better the compression but the worse the quality. Somehow the data in each block is then encoded in the form of some function which doesn't store the exact data hence the data loss (but very high compression ratio). During decompression the blocks are reconstructed from the compressed data. The exact way that works I don't know.
Each frame of video is split into 8x8-pixel blocks. The blocks are then converted from the spatial domain to the frequency domain using discrete cosine transform (DCT). Lots of other stages are involved, including Huffman encoding and differential encoding; all of which help to improve the compression ratio.
Originally posted by michaelab
MPEG2 also uses compression from frame to frame so that parts of the image that don't change don't need to be compressed all over again. Only parts of the image that need to be re-encoded. This is why moving objects are often surrounded by a strange blocky 'halo' where the compression algorithm has tried to detect what changed. Clearly, the more accurately this is done the less artefacts you get.
MPEG1 uses the same technique. So does MPEG-4. In fact, most video compression techniques use motion compensated prediction where movement of pixels or blocks of pixels are encoded as motion vectors. It's essential for getting rid of redundancy and thus reducing the overall bit-rate.
Originally posted by michaelab
There are probably some errors in the above but it's basically how it works. Others with more knowledge please feel free to correct me
Done
Best site for finding out more is
www.mpeg.org