Predictive Coded Pictures

P-pictures are derived from preceding pictures. They refer to the status, which results either from the last I-picture or from preceding P-pictures. Temporally regarded P-pictures may access only to preceding contents, but never go beyond the last I-picture.

In principle two options are available for coding of P-pictures:

Motion compensation:

In the preceding picture a segment is looked for that is as identical as possible. A movement vector refers to this segment and only the difference has to be coded.

Direct Coding:

If no segments can be found providing a sufficient similarity, this segment must be encoded directly in the same manner as an I-picture.

For the quality of coding it is crucial, in which measure similar segments can be found. With a higher accuracy the expenditure for the search increases however exponentially.

It is not necessary that a P-picture refers to the directly preceding picture. Also older P-pictures or the last I-picture may serve as a reference.

 <   ^   >