Motion Estimation Algorithm Analysis and Optimization in H.264/AVC Standard
|School||Zhejiang Normal University|
|Course||Computer Software and Theory|
|Keywords||H.264/AVC Video compression coding Block motion estimation Starting point forecasts Fast Search Algorithm|
H.264 video compression standard for its method is simple, with an estimated effect, and coding and high efficiency is very popular video compression industry's favor. H.264 video coding standard can maintain the same image quality, save up to about half the number of bits than previous coding standards. But due to its standard introduces a variety of new coding techniques, such that the computational complexity of the codec has increased dramatically. Wherein up to 80%, based on the amount of calculation block matching motion estimation operator accounts for the entire video encoding. The high complexity of computation has become the bottleneck of video coding applications in real time. Therefore, low-complexity and high-precision motion estimation algorithm research has become a hot research topic in recent years in the video compression technology in the field. This paper introduces in detail the working principle of the block matching motion estimation, and describes several classic fast motion estimation algorithm characteristics and their advantages and disadvantages. Then, by analyzing the motion characteristics of the video sequence, we propose a parallel search, and multi-template based on the size of the cross template adaptive search motion estimation algorithm. The algorithm makes full use of the characteristics of the video sequence is stationary or quasi-stationary block majority, while using the multiplexing termination strategy, so the algorithm greatly simplifies the search process of motion estimation, H.264 encoding speed. Meanwhile, the multi-template search optimization algorithm used in the adaptive the exercise intensity framing method, to ensure that the motion estimation accuracy and coding efficiency. In order to verify the effectiveness of the algorithm, this algorithm is applied in the reference model, H.264, while a system simulation experiment, and were compared with two classic algorithm (FS of algorithms and UMHexgonS algorithm). Through the analysis of experimental data, the experimental results show that: the algorithm under the premise of no significant change in the image quality and stability of the algorithm, the motion estimation process time-consuming than UMHexgonS operator saves 15.88%. In fast motion estimation, the peak signal-to-noise ratio (PSNR) values, as well as compressed bit rate performance and weighed against the performance of the proposed cross search algorithm based on adaptive template starting point is generally better than UMHexgonS algorithms. At the same time, the proposed algorithm is more robust, more conducive to real-time applications.