Research Of Optimized Image Segmentation Based On Improved Genetic Algoirthm

Image segmentation is the pretreatment of pattern recognition and graphical analysis,it’s the key step of image processing, and it’s also a kind of basic computer visiontechnology, come into notice from researcher from many kinds of field.Genetic algorithm is an optimization method based on random and natural selection.In the last few years it shows great potential in working on complicated optimizationproblem and succeeds industry application. It walks into people’s sight.Genetic algorithm is uncomplicated, robust, parallel and self-adapt. After addedpot-unit, cross-unit, variation-unit and new-unit, genetic algorithm avoids local optimum,increase convergence speed and overall situation searching ability. Genetic algorithm is aparallel algorithm. It has great potential in running speed. Image segmentation iscomplicated and cast much running time. So it in image segmentation field, geneticalgorithm regularly used to determine threshold.In this paper, I present denoising maximum entropy image segmentation on the basisof maximum histogram entropy and2D maximum entropy. The approach is divided intotwo stages: before the segmentation, we first make an image despeckle processing, and inthe process of seeking the threshold, particle swarm optimization genetic algorithm(PSOGA) is used to reduce computation time and improve solution accuracy by combiningthe standard genetic algorithm (GA) with particle swarm optimization (PSO). Throughsegmentation test and comparison, the results obtained by our method are shorttime-consuming and strong anti-noise ability. Computational results show that PSOGA hasgood global optimal search capabilities and faster search speed.I also used parallel calculate ability of OpenCL to increase calculation speed andminimize running-time.