Research On Real-time Video Encoding Optimization

With the development of social applications, real-time video encoding technology is widely used, especially in real-time monitoring. Not just large enterprises need security monitoring, something personal belongings such as private cars and family thing also important. China Telecom use 3G network realization the world's eyes to monitor the family at any time and adopt the mobile video telephone to communication with families. The development of the home network of multimedia services has begun. However, video encoding as the core of video technology, can not meet all kinds of environment, which is the reasons of the high cost of video telephony and hard to promoting. To solve this problem, this paper study of the H.264 which is the most popularly coding technology current in-depth, and analyzes the key technology of H.264. In the end select the motion estimation and integer transform and quantization as the breakthrough point to improve and optimized the speed of encoding . Finally, under the premise of a satisfactory video quality achieve the purpose of speed up the video encoding. The main researches in this dissertation are illustrated as follows:(1) Research of the Principle and methods of Motion Estimation which is the key technologies of H.264 deeply. Study the three types of search algorithm used to carry out in the Official version detailed, at the last selected one of the better algorithm-Unsymmetrical-Cross Muti-Hexagon Search to research and improve. Add the adaptive threshold in the process of the search algorithm, which played a role of early cut-off. And increase an rectangular search step to its five-step search, in most cases it can avoid the time-consuming six polygon search step. Because the motion estimation are more time-consuming process in the encoding, through our work improved the speed of motion estimation, thereby increasing the overall speed of encode.(2) Integer transform and quantization process plays an important role in the whole process of encoding, also a very time-consuming step. This article in depth study and research the integer transform and quantization process which adopted by the H.264, as well as its basic theory -discrete cosine transform knowledge. Through our study and research and studied the algorithm of all-zero block of the decision process, finally joined the all-zero block of the decision process in the whole search process, save the coding and quantization time for the detected zero-block-wide. And based on the algorithm of all-zero block of the decision process,further improvements the threshold selection and search process. Due to the reason of All-zero block relatively large percentage, after the improvement of this article, further shortened the time of encoded.(3) Finally, use the version of the official JM14.0 to validate the proposed algorithm. According to the experiments, the realization of the encoding algorithm is a better picture of the results and demonstration by experiments that the encoding speed has been greatly improved.