An Improved Meta-heuristic Optimization Algorithm And Its Application In Image Processing

Meta-heuristic algorithms have been widely used since they were proposed in the 1960 s as they can effectively reduce the amount of computation and improve the efficiency of optimization.The algorithms imitate all kinds of operation mechanisms in nature and have the characteristics of self-regulation solving the problems like low computational efficiency and poor convergence of traditional optimization algorithms such as Gradient Descent,Newton's method and Conjugate Descent.And the algorithms have good effects in combination optimization,production scheduling and image processing.Image segmentation is an important part of the image processing system,and the segmentation result has a huge impact on the subsequent processing of the image.However,traditional image segmentation often has many problems,such as large amount of data processing,complex function model and so on.In this paper,we focus on meta heuristic algorithm,improve its shortcomings,and apply it to image segmentation.The main work is as follows:(1)In order to solve the problem that the existing one-dimensional K-Entropy algorithm is susceptible to noise and inaccurate segmentation,this paper extends the one-dimensional K-Entropy threshold selection criterion function commonly used in image segmentation to two-dimensional,and gives the definition of two-dimensional K-Entropy and the corresponding segmentation function.The threshold method not only takes into account the value of the pixel itself,but also takes into account the neighborhood information of the pixel.By adjusting the parameter k,the segmentation of different images can be completed well,which has high flexibility and universality.(2)In order to solve the problem that the existing Beetle Antennae Search Algorithm(BAS)is easy to fall into the local optimum when facing the complex nonlinear functions,and the step length and search distance will decline in the later stage of iteration,a learning strategy with strong global exploration ability is proposed: binary discrete Beetle Antennae Search Algorithm(BBAS).(3)Bas is easy to fall into local optimum due to the attenuation of step size at the end of iteration,while BBAS is more random when it runs to the end which is a global search ability and can not converge globally due to its strong global search ability.According to the advantages and disadvantages of the two algorithms,a new Beetle Antennae Search Algorithm(NBAS)is proposed by combining the original bas and BBAS,and using BBAS to assist the original BAS.The algorithm balances local and global search,and effectively makes up for the deficiency that bas is easy to fall into local optimum.The effectiveness of the algorithm is verified by the test on benchmark function.(4)Combining NBAS algorithm and BBAS algorithm with two-dimensional K-Entropy threshold segmentation criterion function respectively,NBAS-KEntropy segmentation algorithm and BBAS-K Entropy segmentation algorithm are proposed.The experimental results on the Berkeley data set,artificial noise-added images and remote sensing images show that the NBAS-KEntropy image segmentation algorithm not only has good anti-noise performance,but also has high accuracy and robustness,and can segment image quickly and effectively.