基于CUDA的红外图像处理算法设计-通信与信息系统专业论文.docxVIP

华 华 中 科 技 大 学 硕 士 学 位 论 文 II II Abstract In recent years, with the tremendous development of the infrared thermal imaging technology, it has been widely used in various fields and has broad prospect. Limitations of infrared device can bring in interfering noise and the imaging machanism can cause the low contrast, the quality of infrared imaging is not such satisfactory. Extensive reseach on infrared image process has been carried out all over the world to solve the problem. Nonuniformity and bad pixel are key factors influcing the quality of infrared images. Accordingly, this paper analyses the production mechanism of nonuniformity and bad pixel, and explored response model of probes. Therefore, in order to meet this request, this thesis designs the infrared image processing system based on CPU+GPU .Using the serial processing of CPU to manage resource and process schedual and using the parallel processing of GPU to process infrared image.As the host processor of infrared image algorithms , GPU can effectively separate CPU. Compared with single CPU processing system, the sytem can effectively decrease the load of CPU and increase the process rate of the whole sytem. Through adopting the CUDA programming scheme, the system realized the modified non-uniformity correction based on scene and adaptive bad pixel correction at upper computer end. These algorithms take fully advantage of theParallel computation ability of GPU,and the upper computer end can accomplish large frame infrared image processing. In the infrared image process system performance test, for 320 × 256 frame rate infrared image, the system real-timely processed the large data of the infrared image .Compared with the deficiency of CPU system, the CPU+GPU system can successfully realize the non uniformity correction based on scenes and adaptive bad pixels correction.The experiment results indicate the effectiveness of algorithms and advancement of this system. Keywords:Infrared image processing system；