| IOS/IECMPEG and ITU-TUCEG joinly launched the newest international video coding standard H.264.With the development of society and technology,people puts forward of different fields, such as Video Conference, Real-time Video Monitoring System,High-definition TV, needs to guarantee higher image quality and compression ratio and make transmission speed faster.Compared with H.263, H.264standard improvements a higher compression ratio, however, frame and Inter-frame prediction algorithms are so complex that cost a lot of time in the encoding process, for instance, the Inter-frame coding takes up60%to80%in the total encoding time.Therefore, how to reduce the algorithm complexity with guaranteeing the quality of image,meet the requirements of different areas and reduce the encoding time becomes many organizations and video encoding amateur research’s hot issues. This article is based on reducing complexity of video coding and optimization algorithm.For the first,introduces the basic knowledge of video coding, a detailed analysis of Intra-frame and Inter-frame prediction theory. In the Intra-frame prediction, some scholars propose a fast coding algorithm based on a large number of zero elements of macro blocks, but it is not applied to the Inter-frame prediction.Therefore, the new algorithm that based on zero macro block applied to the Inter-frame prediction coding is raised.16×16macro block of pixels based on volatility forecasting concept is used in this paper, the main idea is to make the reference frame of the macro block pixel values together, with the value minus the temporary SAD values to generate inequality.When some pixel values which does not meet the formulas will skip it, this can make a prediction mode selection easier, effectively reduce the algorithm complexity and improve the coding efficiency. The origin of the formula refers to SAE algorithm which calculates the whole motion compensation macro block for SAD values in addition to get an inequality, if SAD negative and real numbers are almost equal, it will cause the exclusion pixel values not accurate. Therefore, using the method of16×16pixels of the volatility can avoid the calculation of motion estimation vector value. At the same time, in the new three-step search algorithm combined with the prediction center bias, improve the predictive accuracy and greatly reduce the algorithm complexity.For the experiments showing, in adopting the new three-step algorithm, using the16×16pixels of the volatility can effectively save encoder encoding time, improve the prediction accuracy and PSNR. After a comparison of FS algorithm, old three-step search algorithm and the new three-step search algorithm,the new three-step algorithm performance has improved. Using akiyo、bridge-far、claire、grandma、silent、hall、mobile test sequence, motion estimation time compared to the original three-step search algorithm and FS algorithm reduced by an average of35.77%and6.21times, PSNR(Y) decrease by1.73%and2.52%on average.Three steps of the algorithm’s time under the condition of the PSNR(Y) do not have too big loss reduced, in the case of many video motion smooth image the effect is slowly obvious. Improved three-step search algorithm suitable for similar video conference to real-time demand is higher and the image quality and resolution requirement is not high under the environment of use. |
PDF Full Text Request