H.264 Video Coding Standard Fast Algorithm And Optimization

H.264 video coding standard is based on the MPEG and H.263 video compression developed from the coding standard. Its purpose is to make the coding image used in different network environments with a flexible way, enhance its error resilience capability, achieve to the high-quality images results with the low bit rate.H.264 standard introduces the methods such as the quantization of the DCT integer transform, the 1/4-pixel precision motion vector, the multiple reference frames prediction, deblocking filter, and the new ways of intra-frame prediction method, these methods make the H.264 coding to become more efficiency。The choose of the Intra-frame prediction mode is one of the important problems in affecting the H.264 coding rate. H.264/AVC is the intra prediction in avariety of models from multiple directions, and uses the rate distortion optimization algorithm to select the best mode, As the model has 13 species, which causes a high degree of algorithmic complexity. This article is based on the current analysis of intra prediction mode of fast algorithms situation and proposes a new improved algorithm. By describing the histogram of the image gray distribution information, as a basis judging for the degree flat images, and then proposes the dual-threshold to distinguish the macro blocks which have obvious features, and distinguish the rest of the macro blocks by defining the neighboring models.Firstly, according to use the way of proposing the dual-threshold by the image of the gray degree histogram to judge the large model of the Intra-frame prediction, 16×16 or 4×4. By using the histogram of the image gray distribution information, as a judging for the flat degree of the images, the relatively flat degree is the 16×16 mode, the uneven degree is the 4×4 mode. So propose the dual-threshold to improve the accurancy of the judgement, use the relevant of the macro block mode to make a further judgment if the model is not judged obviously. Secondly, make a further judgment on the base of the large mode judged by the first step. Judge the most probable forecasting model, if it contents the threshold which is setted, and then stop the model searching, if not, then continue the model searching until it finds the optimal forecasting model. Utilize the relevant of the macro block mode to confirm the threshold, and use the SATD of the adjacent models to set the threshold dynamicly.After a great deal of experiments: the algorithm can make the intra-coding efficiency improved between 10%-30%, under the condition of basically maintaining the image quality and encoding performance.After optimizing the H.264's arithmetic, we use the Intel's MMX to optimize the instruction sets of the H.264. Use its instruction sets, which based on the single instruction multithreading technology, to optimize the instructions of the H.264, such as the SAD calculation, the DCT transform, the pixel interpolation, and so on. Experiments show that compared with the C language instructions, the speed of each module is increased through the optimization, and the optimization also reduces the consumption of time by calling the function.