Research On Parallel Computing Of Hevc Transformcoding Based On Epiphany

With the rapid development of network and multimedia technology, the video image transmission need to be stored by the amount of data is also becoming more and more big.So on the premise of the video image definition unchanged, made its storage capacity decline the key to solve the problem of high resolution image in the video transmission.Among them through the transform coding to compress video, is a kind of effective means to solve these problems.At present widely USES the discrete cosine transform to video image compression coding, such as h. 264 coding standard, and in February 2013 launch of a new generation of video coding standard H. 265/HEVC.To solve the problem of high resolution video image transmission, in this paper, the high performance video coding(HEVC) transform coding part has made the improvement, the Epiphany of multi-core processors by parallel computing transform coding of video image compression processing, and one of the most time consuming large part matrix multiplication parallel algorithms are designed, the computational efficiency of transform coding is further improved.Paper main research results:(1) In view of the transform coding’s large amount of calculation, and transform data is two-dimensional, multidimensional data processing need to be in a lot of data dimension reduction processing, the algorithm complexity is larger, reduces the computation efficiency of the phenomenon.In DCT transform coding, for example is deduced the integer transform, due to the transformation matrix are integers, so through the shift operation can save a lot of addition and multiplication, improve the efficiency of calculation.And in the data in a two-dimensional matrix is conducted within each sub-block chunking, so you can more easily through parallel processing.(2) Because there are many large matrix multiplication in transform coding, if by serial achieve its calculation will take up a great deal of system resources, therefore the Epiphany of multi-core processor for parallel computing of matrix multiplication.Through parallel computation is given to implement scalable big Cannon algorithm for matrix multiplication, in order to improve the efficiency of the matrix multiplication calculation.(3) Based on Zynq mixed with 16 nuclear Epiphany architecture, the use of dual Arm A9 architecture to run HEVC encoder based on HM model of ubuntu system.By multi-core hardware platform for parallel computing based on Epiphany HEVC transformation coding for validation.With the result of the experiment and feasibility were analyzed, and finally to the above two parts of multi-core parallel processing is applied to the standard encoder HM model.