초록 ▼

A method and apparatus for motion vector compression is described. Compression is performed in the context of a two-dimensional processing system in which two-dimensional frames are each encoded into minimum-error motion vectors. For each frame, the minimum-error motion vectors are compressed by his

A method and apparatus for motion vector compression is described. Compression is performed in the context of a two-dimensional processing system in which two-dimensional frames are each encoded into minimum-error motion vectors. For each frame, the minimum-error motion vectors are compressed by histogramming at least one minimum-error motion vector to generate at least one histogram peak. N most dominant peaks represent the frequency of occurrence of N most dominant motion vectors. Compression further requires that each minimum-error motion vector is mapped to one of the N most dominant motion vectors. Before undergoing histogramming, each minimum-error motion vector may be mapped to a null motion vector (0, 0) if the minimum-error motion vector does not provide a significant improvement over the frame-to-frame correlation error associated with the null motion vector for the frame. In one embodiment, the mapping of the minimum-error motion vectors to the N most dominant motion vectors may be carried out by mapping a minimum-error motion vector to a corresponding dominant motion vector, where the frame-to-frame correlation error associated with the corresponding dominant motion vector is less than the frame-to-frame correlation error associated with all other dominant motion vectors associated with the minimum-error motion vector.

대표청구항 ▼

[ What is claimed is:] [1.] In a system in which frames are each encoded into a plurality of minimum-error motion vectors, each motion vector has a frequency of occurrence within a frame, wherein each frame is divided into blocks, a method for compressing the minimum-error motion vectors comprising:

[ What is claimed is:] [1.] In a system in which frames are each encoded into a plurality of minimum-error motion vectors, each motion vector has a frequency of occurrence within a frame, wherein each frame is divided into blocks, a method for compressing the minimum-error motion vectors comprising:(1) creating a histogram of at least two minimum-error motion vectors to generate at least two histogram peaks; and(2) mapping other motion vectors using error-minimization criteria to one of the at least two histogram peaks, the other motion vectors being located in a spatial neighborhood around the one of the at least two histogram peaks, the spatial neighborhood having a boundary determined by a characteristic of the at least two histogram peaks, the characteristic being one of a bisector intersection between the at least two histogram peaks and a weighted dominance such that a more dominant peak in the at least two histogram peaks has a larger size than a less dominant peak in the at least two dominant peaks.