Motion estimation plays a key role in motion picture compression, because accuracy in finding motion vector greatly affects the quality of the compressed images. For the real time operation, motion vector searching speed is also of great concern. So, the conventional searching algorithms are focused on the reduction of searching speed. The proposed block matching algorithm, however, is mainly aimed at finding the motion vector that gives higher compression rate.
In the proposed algorithm, a Macroblock is divided into four Blocks, and each Block is further divided into four Sub-blocks. Noting the fact that the compression is done by block basis while motion is searched by Macroblock basis, the algorithm searches for the motion vector that gives the best match of some of Sub-blocks in the Macroblock. To further reduce the bit rate, matching error of vector (0,0) is first calculated and if it is less than predetermined threshold we take it as the motion vector of the Macroblock. Also, the sums of 16 pixels are used as basic elements of matching criterion to reduce computation time.
The computational complexity of the proposed algorithm is compared to those of the conventional algorithms. The performance of the proposed algorithm in terms of PSNR and bit rate is also compared with other algorithms using 5 video sequences of 50 frames respectively.