An improvement of bottom-up variable-sized block matching technique for video compression

Block matching can reduce temporal redundancies. It is an important scheme in video compression. So far, there have been many methods proposed for this scheme, such as fixed-sized block matching (FSBM), top-down variable-sized block matching (TDVSBM), and bottom-up variable-sized block matching (BUVSBM). In these methods, the blocks in each frame of a video must be square, and that is why some objects in the frame cannot be described correctly. To remedy this problem, we propose a new block matching method. In this new method, the blocks in each frame of a video do not need to be square. Moreover, all kinds of combinations of small blocks are taken into consideration. Because our block shapes are dynamic, all objects in the frame can be described clearly, and, under the same number of blocks in each frame, our new method can generate better image quality than FSBM, TDVSBM, and BUVSBM. This phenomenon is proven in our experimental results.