Design Of Infrared Imaging Processing Algorithm Based On CUDA

In recent years, with the tremendous development of the infrared thermal imagingtechnology, it has been widely used in various fields and has broad prospect. Limitations ofinfrared device can bring in interfering noise and the imaging machanism can cause the lowcontrast, the quality of infrared imaging is not such satisfactory. Extensive reseach on infraredimage 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, andexplored response model of probes.Therefore, in order to meet this request, this thesis designs the infrared image processingsystem based on CPU+GPU.Using the serial processing of CPU to manage resource andprocess schedual and using the parallel processing of GPU to process infrared image.As the hostprocessor of infrared image algorithms, GPU can effectively separate CPU. Compared withsingle CPU processing system, the sytem can effectively decrease the load of CPU and increasethe process rate of the whole sytem.Through adopting the CUDA programming scheme, the system realized the modifiednon-uniformity correction based on scene and adaptive bad pixel correction at upper computerend. These algorithms take fully advantage of theParallel computation ability of GPU,and theupper computer end can accomplish large frame infrared image processing.In the infrared image process system performance test, for320×256frame rate infraredimage, the system real-timely processed the large data of the infrared image.Compared with thedeficiency of CPU system, the CPU+GPU system can successfully realize the non uniformitycorrection based on scenes and adaptive bad pixels correction.The experiment results indicatethe effectiveness of algorithms and advancement of this system.