Application Research Of Image Segmentation Based On Hybrid Heuristic Algorithm

Image segmentation is the course of dividing an image into several subregions that do not overlap each other and withdrawing out the meaningful regions.Image segmentation is the vital measure for moving from image processing to image character extraction and recognition,and its in quality will influence the follow-up image analysis and pattern recognition.Heuristic algorithms are widely used to deal with complex optimization problems in various fields.Image segmentation can be seen as finding the best parameter values to segment images in a complex data space,and heuristic algorithms have unique advantages in solving complex nonlinear multimodal function optimization problems.The goal of the clustering algorithm is to classify all pixels in an image according to the similarity criterion,so that pixels with greater similarity are divided into the same class,and unsimilar pixels are divided into different classes.K-means cluster segmentation is also one of the most commonly used methods in cluster segmentation,and the effect of Threshold segmentation is an efficacious and practice segmentation method,the mechanism of which is to seek out the most optimum threshold value based on a certain criterion,and then segment the image depending on the gained threshold value.The computational sophistication of the threshold segmentation algorithm grows rapidly as the dimension of the image information or the number of threshold values selected are increased,which considerably increases the time required for computation and to certain extent restricts the range of use of the algorithm.To this end,two new heuristic algorithms are improved and applied to image cluster segmentation and image threshold segmentation,respectively,as follows:(1)Aiming at the shortcomings of traditional K-means image segmentation that is random and easy to fall into local optimality,a K-means image segmentation based on Halton sequence improved manta ray algorithm(HMRFO)is proposed,which uses halton sequence to initialize the population so that the individual position is fully uniform.Refractive reverse learning is introduced to improve the global search capability of the algorithm.Finally,a new Gaussian variation strategy is introduced to reduce the probability of the algorithm falling into the local optimum.The effectiveness and feasibility of HMRFO are verified by standard test function,and compared with whale optimization algorithm,particle swarm algorithm and manta ray algorithm,and the image segmentation optimized by K-means can be proved to be better than other algorithms.(2)Aiming at the problems of large computation and long time of computation in conventional multi-threshold image segmentation methodology,an image segmentation method with multi-strategy improved butterfly optimization(CLBOA)and Tsallis entropy method with maximum entropy is proposed.The method adopts Cubic chaotic mapping to initialize the population,enhance the diversity of the population and the diversity of the solution,and enhance the capacity of global search;then adjusts the position update method of the algorithm by introducing Levy flight strategy;and perturbs the optimal solution through the Corsi variation;finally combines with Tsallis entropy threshold segmentation method to search the optimal threshold for image segmentation through multiple iterations.In order to test the algorithmic performance of the proposed method,the effectiveness and feasibility of the method are verified on a benchmark test function.Images from the Berkeley Gallery were picked for image segmentation trials and were compared with five algorithms for analysis.The results demonstrate that the method outperforms the comparison algorithms in terms of cut accuracy,computation time and convergence,and can rapidly and effectively solve the multi-threshold segmentation problem of complex multi-target images.