Developer Home Contents Search Feedback Support Intel(r)

Use of Bidirectional Prediction

Question: What is bidirectional prediction used for?
Compression can be of two kinds: intraframe and interframe. Intraframe compression compresses the data within one frame relative only to itself. Key frames are compressed with intraframe compression because they must reconstruct an entire image without reference to other frames.

Interframe compression compresses the data in one frame relative to others. These relative frames are called delta frames. If the information in delta frames is encoded relative only to previous frames, then that is unidirectional compression--the codec always looks back in order to decompress the video image.

Bidirectional prediction is a form of compression in which the codec uses information not only from frames that have already been decompressed, but also from frames yet to come. The codec looks in two directions: ahead as well as back. This helps avoid large spikes in data rate caused by scene changes or fast movement, improving image quality. Bidirectional prediction adds a certain amount of playback overhead; for a clip without much movement, such as a talking head, it may not significantly improve visual quality.



* Legal Information © 1998 Intel Corporation