METHOD AND DEVICE FOR FORWARD MULTIPLE-HYPOTHESIS ENCODING/DECODING OF IMAGE BLOCK

摘要

The present disclosure provides a forward multiple-hypothesis encoding/decoding method of an image block. In this method, motion information of a current block including two forward reference frame indexes and two motion vectors are selectively encoded into a bit-stream based on a current encoding mode. The motion information which is not encoded into the bit-stream is derived at a decoder by a predefined rule. This method sufficiently utilizes correlation within motion information of an image block and correlation between the motion information of the current image block and that of an adjacent image block.

主权项

1. A forward multiple-hypothesis encoding method, comprising:
(1) for an image block A, obtaining a reference image block B and a reference image block C by performing a motion searching in a reference queue, and obtaining a prediction image block P of the image block A using a weighted sum of the reference image blocks B and C; (2) obtaining a reference frame Fb where the reference image block B is located and a motion vector MVb based on a position of the reference image block B, obtaining a reference frame Fc where the reference image block C is located and a motion vector MVc based on a position of the reference image block C, obtaining a reference frame index RFb based on a position of the reference frame Fb in the reference queue, and obtaining a reference frame index RFc based on a position of the reference frame Fc in the reference queue; (3) writing information of an encoding mode selected by an encoder into a bit-stream; and (4) encoding some of RFb, RFc, MVb and MVc into the bit-stream based on the selected encoding mode.