| With the development of information technology, human demand for more richness of information, multimedia information has become the main carrier to obtain information. Visual information has intuitive and vivid features, the amount of information the human vision get is about 70% of the total amount of information, so video information plays an important role in multimedia information. But the redundancy of video data is great, video coding research has become a hot topic in the field of information technology.Motion estimation as a key video compression technology is a very important part in the video compression encoding system. It directly affects the quality and efficiency of video data compression encoding. Motion estimation is also the most complex and time-consuming part in coding operation, according to statistics the computational complexity of motion estimation accounts for more than 50% of the entire video data compression encoding system. Therefore, the studying of effective motion estimation algorithm to improve the efficiency of video data compression encoding has a very important practical significances, it is an efficient solution to the current video data compression problem. In a variety of motion estimation algorithms, as block matching algorithm is simple and easy hardware implementation, it has been widely used and researched.This paper introduces some video compression coding techniques and the principle of a series of international video coding standard; also do a detailed introduction to H.264 video coding standard. Then the paper introduces the basic principles of motion estimation, macro-block matching criteria and some typical motion estimation algorithm. The realization of these kinds of typical motion estimation algorithm on JM86 model has laid a theoretical basis for further optimization and design algorithm.On this basis, we proposed two new motion estimation algorithms. One is improved UMHexagonS algorithm, the other is the cross– octagon searching algorithm. Improved the original UMHexagonS on three aspects: according to the characteristics of motion vector distribution probability and the macro block partitioning in H.264 video coding standard, the proposed algorithm replaced the 5×5 spiral full search of UMHexagonS algorithm by a small non-symmetrical cross template searching; replacing the hexagon searching template by octagon template in order to get better performance, and add the horizontal-type hexagonal search and vertical-type hexagonal search template in the proposed algorithm. We have simulated the algorithm on JM86 model. From the simulation results can be seen that compared with the UMHexagonS algorithm, the new improved algorithm ensuring the quality of coded did not change, but reduces the time of motion estimation about 30%. Thus, the new improved algorithm is an effective motion estimation algorithm and contributed to real-time encoding applications.Cross-Octagon searching algorithm takes full advantage of the center-biased characteristics of motion vector and spatial-temporal in video sequences, motion types are divided into fast motion type, mild motion type and slow motion type, the method of selecting the starting point is the sum of absolute method. Combining with cross-shaped template and octagon search template, adaptively selected the search strategies and search starting point according to movement type, at the same time set the threshold directly suspended the search for stationary blocks, threshold setting does not use a fixed value as before, but according to the spatial correlation of macro block calculates the threshold to improve the searching accuracy. Stationary block suspended technology avoiding searching redundant searching points, raised the efficiency of the algorithm. The experimental results show that the searching speed of the Cross-Octagon algorithm is better than the existing fast motion estimation algorithm, the search accuracy of the proposed algorithm close to UMHexagons algorithm. |
PDF Full Text Request