| 摘要 |
A sequence of images is compressed. A reference image (30) and a first image (32) of the sequence are received and background image data is stored in a background buffer memory (14). Motion vectors of respective image regions (34a,b, 36a,b) from the reference image (30) and the background image to the first image of the sequence are determined. The motion vectors are used to identify an occluded region (39) in the reference image (30) between image regions (36a,b) with converging motion vectors. Image data derived from the occluded region of the reference image is selectively stored in a background buffer memory (14). Later, when a second image of the sequence is received the stored image data in the background image memory may be used to determine a motion vector for encoding the second image. Preferably, the background image data is motion compensated as data from progressively more images is stored. Preferably, background image data is also accumulated for use during decoding, but alternatively the motion vectors are applied to one of series of foreground reference images during decoding. In this case, information about the source of the background is kept during encoding, for transmission with the motion vectors. |
