Fragmented reference in temporal compression for video coding

摘要: In general, this disclosure describes techniques for encoding and decoding sequences of video frames using fragmentary reference pictures. The presents modified temporal compression based on fragmented references rather than complete a typical sequence frames, only portion (i.e., tile) each frame includes moving objects. Moreover, in frame, the objects tend to be confined specific areas that are common among frames. As described herein, such motion identified. Pictures then extracted from identified Because these pictures may represent portions refers as "fragments." It is fragments used generating predicted during compensation process, entire frame.