The motion estimation and compensation techniques are widely used in video coding applications because of their capability to reduce temporal redundancies residing in successive frames. As a motion estimator, the block-matching algorithm(BMA) is widely used because of its simplicity and regularity. But the required computations for the BMA are so large that it is not possible to perform the BMA in a timely fashion. Therefore, developments of VLSI architectures or new fast search algorithms are indispensable. In this thesis, we discuss the adjustment of search window for the BMA to reduce the computational complexity, as a method for fast search algorithms in low bit-rate video coding.
We propose a dynamic adjustment of the search window with fixed size of blocks(DASWF) for the BMA to reduce the computational complexity of full search algorithm(FSA). The method uses block similarities and displaced block differences to adaptively adjust the size of search window. The technique can be easily applied to the FSA and several fast search algorithms to get more efficiency. The experimental results have shown that the mean square error(MSE) performance and the reduction of the number of search points with the proposed scheme are better than those of previous works.
The dynamic adjustment of the search window with variable size of blocks(DASWV) is also presented to improve the accuracy of the motion estimation. As similar to the DASWF, this method uses the segmentation information of current frame to divide the frame into variable size of blocks. The size of search window for each block is determined by using the correlation of block similarities of adjacent blocks and their motion vectors. We also proposed a new method for search window adjustment using the spatio-temporal correlations, especially block classification information. The method exploits the correlation of successive video frames and adjusts the size of search area depending on the displaced block differen...