Parallel Blind Image Restoration Technology

The image processing time is very important for some applications, especially for military uses; therefore real time image processing has become a research hot topic recently. However, the traditional real time image processing systems are always designed for special application, hardware and algorithm, and have to be redesigned whenever the image operation changes, which obstructs its application greatly. Depending on the parallelizability of data in image processing, this article establishes a parallel image processing system constituted by PC Cluster connected by the Ethernet, on which most of operations of image processing can be run, and because the wide application of the PC and Ethernet, this system has a promising future. The main research work can be summed up as follow:Establish a military parallel image processing system, on which most operations of image processing can be run.After a comprehensive survey in the parallel computer architectures, this article finally chooses the PC Cluster connected by the Ethernet as the hardware to build the system, and then design a master-slave software model for it. Because of the parallelizability of data in image processing, this system can reach a high efficiency, low expense, good expansibility and is easy to be developed.Figure out the conditions of the least execution time.Load balancing is one of the most important questions in parallel processing. This article establishes an analyzing model for the above system, and answer the question of the least execution time. The experiments followed prove the validity.Design and complete a parallel blind image restoration algorithmWhat's more, this article also designs and complete a parallel blind image restoration algorithm depending on the system provided above. Image restoration is one of the most important branches in image processing, and has been widely applied in lots of fields, such as astronomy, medicine, remote sensing and industry. The traditional image restoration requires the point spread function of the blur, but a new technology of blind image processing don't. Obviously, the blind image restoration can be more useful in real applications. This article analyses the existing algorithms, then designs a parallel edition for the NAS-RIF algorithm, which is sure to convergence and likely to have high speed. The experiments show that the parallel algorithm has high efficiency.