Research And Implementation On Super Resolution Of Miture Resolution Image Base On GPU

GPU is a graphic processing unit, due to its excellent parallel data processing capabilities, in recent years more and more used in general-purpose scientific computing. CUDA is a GPU programming platform and launched by NVIDIA, it has developed a very simple process and programming style, using multi-thread programming process, made programming efficiency very highly. The platform leverages the powerful GPU computing capabilities, while the requirement for programmer is not very high, so it has been widely used in compute-intensive applications. The algorithm of mixture resolution image deals with the low resolution image with the correlation of images and information of high resolution image, and restores the original resolution. The main steps and key technologie include the transformation of images of different point of view, projection of images, filling the projection imagesIn the huge amount of data processing, the algorithm involves a large number of parallel computing, we will use the powerful parallel processing capability of GPU to improve the speed of the algorithm.This paper firstly describes the CUDA programming principle, and summarizes the classic super resolution method, and then discusses the super resolution principle of mixture resolution image and its key technologie. On the basis of the existing crack repair methods, we select the non-uniform sampling image recovery, and simulate on the MATLAB, realize the algorithm on the GPU.On the basis of the super resolution of mixture resolution image, the paper further study the optimization strategy of parallel computing, including the Optimization of hardware optimization according to the characteristics of the GPU and algorithm Optimization, further improve the speed of the algorithm, further improve the efficiency of the algorithm. Compared with the simulation results on MATLAB and realization on GPU before optimization comparison, it can be concluded from the experimental results that the super resolution method of mixture resolution image based on GPU is efficient in time consumption without image quality degradation.