Method and system for providing bi-directionally predicted video coding

摘要

A method for providing video coding is disclosed. The method includes encoding at least one segment of a B picture using motion estimation based prediction. During the encoding step, a determination is made as to whether a number of motion vectors corresponding to a number of already constructed reference pictures of the B picture are acceptable for use in a direct bi-directional prediction mode without further refinement. The motion vectors are refined in response to the determination that the motion vectors are not acceptable. A mode is then selected to be used for coding the segment of the B picture. The mode is selected from a number of modes including a temporal direct mode and a number of standard modes including forward, backward and bi-directional predictive modes. The method further includes decoding the segment of the B picture in an encoded bit stream. During the decoding step, entropy decoding of the encoded bit stream is performed to recreate residual data. The mode selected for coding the segment of the B picture is identified. In response to the identification of the temporal direct mode as the selected mode, local motion estimation refinement is performed using the already constructed reference pictures of the B picture. The local motion estimation refinement is identical to the motion estimation refinement performed to refine the motion vectors and the local motion estimation refinement generates a number of local motion vectors. Inverse quantization and inverse transformation are performed with respect to the residual data. Motion estimation based compensation is subsequently performed with respect to the residual data using the local motion vectors. The B picture is then constructed using a number of decoded segments relating to the B picture.