Research And Implementation Of Parallel Video Encoding And H.265Standards-based

HEVC following the launch of a new generation of H.264/AVC video coding standard, bychanging the way the image is divided using the new coding techniques and other methods so thatthe same image quality in ensuring the premise of coding efficiency doubled. Price to pay is asubstantial increase in complexity of the algorithm, making the current hardware platform isdifficult to achieve real-time encoding of HD video. On the other hand, GPU than Moore gainedrecent developments in the field of non-graphics problems can also be accelerated GPU processing.NVIDIA introduced the CUDA programming model makes the GPU more generic applicationdevelopment simple and flexible.For time-consuming coding and other unfavorable factors HEVC produced paper CUDAprogramming model studied in detail the principles and HEVC video encoding technology, basedon intra prediction, transform and quantization and inter prediction of three modules were optimized.Research Working Paper mainly reflected in:(1) In the intra prediction, intra prediction we reduce the data dependencies between the pixels,the image quality is not great losses realized case of parallel processing in the intra predictionCUDA.(2) At the time of the prediction residual transform and quantization, because the datadependencies between small, it can transform and quantization direct parallel optimization, we areusing the global memory and shared memory to achieve parallelism, thus optimizing the parallelrate.(3) Inter prediction is the most time-consuming process of encoding stage, we estimate that thesearch algorithm by selecting the appropriate motion, parallel processing of the motion estimationprocess and the absolute error (SAD) value, and predictive value of the combined SAD processingunit, reducing the overall amount of the SAD value calculation, thereby increasing the codingefficiency.The simulation shows that H.265video encoding optimized acceleration performance on thecorresponding increase. However, due to limitations of hardware conditions, the acceleration is notso high than imaginition.