Research On Image Segmentation Based On Intelligent Optimization Algorithms

Image threshold segmentation plays an important role in the field of image segmentation.The current research results of image threshold segmentation using intelligent optimization algorithms are significant.Since the types of images are all-inclusive and the source of the images is varied,the problem of image threshold segmentation becomes more and more complicated.The existing algorithms still have many deficiencies.They can not meet the actual requirements in terms of time complexity,convergence and so on.In general,a single algorithm can only handle single-type images.And most existing algorithms cannot adaptively select the segmentation parameter.For these deficiencies,this paper proposes two algorithms to improve them.Firstly,for the problems that image threshold segmentation using generalized fuzzy entropy can not automatically select optimal parameters and the complexity of the algorithm is relatively high,the adaptive differential evolution method for selecting the parameters of the generalized fuzzy entropy image threshold segmentation is proposed.The optimization algorithm used to select the optimization parameters adopts the adaptive differential evolution algorithm.The proposed algorithm has the same basic principle steps as the basic differential evolution algorithm.However,the adaptive mutation operator is introduced in this paper to replace the mutation operator in the basic differential evolution algorithm.And the cross-probabilistic self-adaptive functions have also been designed to dynamically change the value of crossover probability CR through evolution iteration steps.The newly proposed algorithm can dynamically change the size of mutation operator and the adaptive function value of crossover probability at different sta ges of evolution according to different types of images.The proposed algorithm can greatly reduce the algorithm time complexity and achieve the adaptive selection of the segmentation parameters.In order to verify the performance of the algorithm,the proposed method is compared with two existing advanced algorithms through a large number of simulation experiments.The experimental results show that the results obtained by the proposed algorithm has lower segmentation error rate than the existing advanced image threshold segmentation algorithms in most cases,and has less background information and clearer target information.Then for the problems that most current algorithms of solving image optimal threshold can not adaptively select parameters and the types of processed images are relatively unitary,the generalized fuzzy entropy image threshold segmentation algorithm based on double adaptive ant colony algorithm is proposed.The algorithm adopts a dual-adaptive mechanism to automatically select generalized fuzzy entropy parameters according to different conditions.The two adaptive mechanisms are initial pheromone concentration adaptation and global update rule adaptation.The initial pheromone concentration adaptation mechanism is used to guide ants to move in the optimal direction and suppress ants to select uncorrelated paths.The global update rule adaptation mechanism is used to improve the convergence of the traditional algorithm.The dual adaptive mechanism can avoid blind search when ants select the path,which greatly improves the speed of solving the global optimal solution.In order to verify the effectiveness of the algorithm,the proposed method is compared with existing four advanced segmentation algorithms through a large number of simulation experiments.Due to the adoption of the adaptive mechanism,the experimental results show that the results obtained by the proposed algorithm can deal with more image types under the same conditions.In addition,the segmentation parameters are adaptively determined by global optimization,which overcomes the time-consuming disadvantage of exhaustive search.The proposed algorithms greatly improve the stability and convergence speed and is very suitable for solving the image threshold segmentation problem.Finally,the work of this paper is summarized and the specific problems to be studied in the future are proposed.