MPEG: A Video Compression Standard for Multimedia Application
by D. Le Gall
Presented by Terence Ho
Summary
In 1988 the MPEG committee had to come up with a video/audio compression standard in two years time. It has to be a standard that is efficient enough that can be used in a variety of ways, and convinces most industry vendors and users to use this standard.
The paper only talks about the video compression of the MPEG-1 only. The video compression takes advantage of intra-frame compression and inter-frame compression techniques to achieve a VHS-like quality with 1.2Mbits/s of bit rate. Intra-frame compression employs DCT functions and quantization on each 8x8 pixel box, which is very similar to the way JPEG does its compression. Intra-frame makes use of IPB-frames and motion vectors to eliminate inter-frame redundancy.
Discussions
There was not very much time for discussions, but a few points were discussed:
- B-Frames are almost never used in live videos(e.g. live broadcast, video conferencing) because it delays encoding.
- MPEG-1 does not perform that well in real situation especially on inter-laced video.
- MPEG-2 improves on inter-laced video and uses half pixel on motion vectors, and in turn achieving double the quality with the same bit-rate.
- MPEG committee is working on MPEG-21 currently
- Error compensation in the motion compression in the presentation is not employed by MPEG-1 nor MPEG-2. MPEG-4 may not have that feature neither, but it does show the short coming of using motion vectors only.
- MPEG-1 video has a compress ratio of about 100 to 1, so it is quite substantial. Such compression ratio is only achievable by throwing some data away, thus the quality of MPEG-1 is only comparable to VHS quality.
- MPEG-1 is only used widely in VCDs, which is a standard to record video on CDs and is widely popular in Asia.
References
included on the last slide on the presentation
Presentation slides