OpenCL-Based Video Jitter Acceleration Method

Due to the rapid increase in the popularity of handheld devices and network performance, a lot of shooting handheld devices and the field of real-time monitoring envisioned has become a reality, but because of physical hardware as well as some personal reasons, there will always be some video jitter motion noise. Jitter function in today more and more prominent role in the new hardware platform, there is often a direct support for some hardware debounce. However, due to the requirements of real-time and user experience, CPU processing video performance often can not achieve the real-time requirements, thus becoming the bottleneck problem in some applications. Modern computer graphics on the parallelism advantage, more and more to help the CPU to complete some computing tasks, while the most prominent is the binary matrix operations reflected when a huge advantage in terms of graphics and image processingIn this paper, now a classic jitter algorithm on the AMD platform GPU to achieve improved performance by optimizing the GPU with respect to the advantage of the CPU in the field of video processing. Finally, based on the test results of the comparison of the three sets of experimental data, the comparison of the CPU and GPU platform, the GPU with respect to the traditional CPU for video processing advantage. The latter two groups are two of the more important points of GPU optimization process, in the process of memory access global variables for the number of visits for the algorithm optimization, as well as in the number of parallel data block selected for the optimization algorithm. GPU in this particular area of the video compared to the advantage of the CPU, as well as in the the popular parallel platform above video optimization algorithm should be noted that some of the problems.