Performance Lossless Optimization of Image Interpolation in H.264/AVC

The motion estimation in the newest international video coding standard H.264/AVC employs many advanced techniques including multiple reference frame selection, flexible macroblock partitioning, and 1/4-pel accuracy motion vectors, etc. Although these techniques improve the coding efficiency dramatically, they also issue heavy computational burden to H.264/AVC codec. Among which the high resolution image interpolation for multiple reference frame is one of the most computational intensive module, therefore requiring more emphasis to be improved. In this paper an optimization scheme of image interpolation is proposed. Simulation results show that this scheme is quite effective and could result in at most 14.4 times calculation speed accelerating. And it is worthy to note that no performance lost occurs during the optimization process. Further results also show that when combined with efforts in other modules, it can bring our H.264 codec to be easily used in real time video applications with CIF requirement