Research On Parallel Image Processing

With the development of parallel processing technology, parallel computing is constantly applied to many engineering fields outside the scientific computing, such as image processing, video compression, fluid mechanics, remote sensing technology, biometrics, etc. Among them, parallel computing technology have played an important role in the traditional areas of image processing. However, parallel processing technology in the development of image processing is also facing many challenges. The most important is how to improve the comprehensive ability to solve practical issues, including the final solution of the complex issue and acceleration.The benefits of parallel processing technology used in many imageprojects mostly depend on the actual complexity of the application and the affordability of execution units for the system cost.This article mainly describes the accelerated solution of two specific images projects by using parallel technology. By modifying theserial algorithm and changing the calculation mode with available resources, we can highlyimprove the operational efficiencyand drastically reduce thecomputation time of the whole project.There are two kinds of accelerated methods we used in this article. One is the multi-CPUdistributed parallel computing technology and the other is GPU’s high-performance parallel computing technology. By analyzing the serial algorithm and specifying a parallel reduction according to the parallel processing rules with efficient hardware resources, we can finally shorten the calculation time on the corner matching and the movement of variable lattice models, including the color interpolation. The volumetric3-D active image project can real-time display due to the parallel processing. MATLAB and Visual C++are the two main platforms we used in the research. And by using the following four tools:MATLAB Parallel Computing Toolbox, Jacket, CUDA and DirectCompute, we accelerate the two specific image projects.And by comparing the results we can know which solution gets the best performance.Finally, we give the results of the four acceleration solutions and it shows that parallel processing technology dose help improve the computing property with accurate data and less-time consuming.