H.264 is the upcoming international standard for video coding which can provide superior rate-distortion performance over the previous H.263 or MPEG-4 standards. However, the computational complexity of H.264 is much higher than that of MPEG-4 especially for the motion estimation process. Two new motion estimation features have been included in H.264 standard - multiple reference frame motion estimation and multiple block size motion estimation. Although superior performance can be obtained by enabling these new motion estimation functions, it can consume up to 90% of the total encoding time. Therefore, reducing the computation complexity of Multi-Frame and Multi-Block motion estimation is much needed....[ Read more ]

H.264 is the upcoming international standard for video coding which can provide superior rate-distortion performance over the previous H.263 or MPEG-4 standards. However, the computational complexity of H.264 is much higher than that of MPEG-4 especially for the motion estimation process. Two new motion estimation features have been included in H.264 standard - multiple reference frame motion estimation and multiple block size motion estimation. Although superior performance can be obtained by enabling these new motion estimation functions, it can consume up to 90% of the total encoding time. Therefore, reducing the computation complexity of Multi-Frame and Multi-Block motion estimation is much needed.

In this thesis, a fast multi-frame selection algorithm and two fast multi-block motion estimation algorithms are proposed to efficiently reduce the computational complexity over the full search method. The proposed algorithms essentially reduce computation by skipping motion estimation for areas with no performance gain. Simulation results show that the proposed Fast Multi-Frame and Fast Multi-Block Motion Estimation can reduce more than half of the original computation while maintaining similar PSNR and bitrate performance.