| With the development of the computer network and electronic technology, multimedia communication which has replaced single media communication is changing all people's life greatly. Because the video information is the most important ingredient of the multimedia information, the digital video encoding technology has been the core and hotspot of the multimedia application. Many neonatal technologies, such as multimedia messaging service, remote education, video conference, video request, digital TV have made the life more and more rich and colorful and it is full of individuation and personal friendliness. Because of the great data information, the video can hardly be transmitted by the available narrowband communication network now without data compression, for example, PSTN network, mobile communication network and so on. Therefore, the research of the video information compression and encoding has become one of the focuses of the IT field.The VCEG of the ITU-T and the MPEG of the ISO/IEC have established H.26x series and MPEG-x series video standard respectively since 1980s. For improving the encoding efficiency further, increasing the video adaptability of the diverse network state, realizing the real time processing and communication of the video at lower rate, those two organizations have cooperated and developed H.264 which is the new generation video compression encoding standard.Like the former standards, the basic algorithm of H.264 is the hybrid encoding mode which is composed of the image prediction and the relativity of the frames coding. As the newest video compression standard, it has many new characteristics in the technology. H.264 has gained the better compression effect of video than other standard by these new characteristics. For example, it consume less 50% bit-rate than H.263 at all rates and get much clearer image at high-rate. But the outstanding encoding performance costs the complicated calculation. In order to increase the encoding efficiency, H.264 adopts the motion prediction algorithm which is based the multiple reference frames, 7 different motion compensation block sizes mode and the rate-distortion (RD) optimization. To achieve as the highest encoding efficiency as possible, each mode and reference frame computed in the coding, and RD optimization is employed to compute the RD Cost. Set the minimum RD Cost's mode and reference frame as the current block's encoding parameter. This process cost most of the coding time.The motion estimation algorithm of H.264 is not only the key of improving coding efficiency but also the bottleneck of the real-time promotion. Therefore, considering both the real-time and the coding efficiency, this thesis presents a new fast motion estimation algorithm of H.264 which is based on the adaptive setting reference frame and the spatial and temporal correlations.In this paper, we first introduce the theory and the technique characteristics of MPEG-x and H.26x standards. Then we study the syntax structure of the H.264. Later, we analyze the key technology of the motion estimation of H.264. Subsequently, we study and research the idea and theory of the motion estimation and compensation. By the simulation experiment, we analyze the complexity of the complicated motion estimation algorithm. We find out that the multiple reference frames and the 7 sizes modes of block matching have taken most of the time of encoding.This paper presents a fast motion estimation algorithm which can adjust number of reference frame adaptively and set block matching mode quickly. Basis on the theory that the first and second frame before the current frame are the best reference frames in all probability, according to the comparison of the second order moments of the motion vector which belong to the two P frames before the current frame, we judge the trend of the video sequence movement and set the best reference frame adaptively and quickly. If the video sequences change smoothly, we set the number of reference frame is 1. Otherwise, we set the number of reference frame is 2 and increase the motion vector search range to enhancing the spatial and temporal area. In this way, we simplify the complicated motion estimation process. When the video sequences change smoothly, we can set the mode of block matching quickly if there are the high spatial and temporal correlations around the current macroblock. So we needn't calculate all the mode costs, but can choose the mode quickly.Basis on the detail analyses of the motion estimation algorithm of H.264 and the sufficient experiments on the JM86, considering the time of motion estimation , PSNR and the encoding efficiency in balance, the advanced algorithm which is presented by this paper perform better than the original algorithm of H.264. It can improving the real-time encoding and provide the new idea and method for motion estimation optimization of the video encoding. |
PDF Full Text Request