Improved Jellyfish Search Algorithm And Its Application In Image Segmention

Multi-threshold image segmentation is a simple,efficient and easy to implement technique.The selection of the threshold can determine the quality of image segmentation.Among the common methods of threshold selection,the threshold selection method based on meta-heuristic algorithm has the advantages of simplicity and ease of implementation,and has become an important threshold selection technique.The Jellyfish Search（JS）algorithm is a new algorithm proposed by Chou et al.in 2020 based on the behavior of jellyfish searching for food.The strong search capability of the algorithm makes it a potentially excellent heuristic algorithm for solving optimization problems.However,the JS algorithm still has some shortcomings.As a consequence,the main research of this paper is to improve the performance of the basic JS algorithm and to use the improved algorithm for selecting the best thresholds for image segmentation.The specific work is as follows:（1）Aiming at problem of solving the multi-threshold image segmentation which computational cost of traditional image segmentation increases exponentially as the segmentation level increases,this paper proposes a multi-threshold image segmentation method based on an improved double-population jellyfish search（IDPJS）algorithm.Firstly,initialize two jellyfish populations P₁and P₂.Perform basic JS algorithm.The combined mutation strategy is introduced into P₁,and the two populations share information by interactive learning to improve the convergence speed of the algorithm.The dynamic opposite learning strategy is used for the current best solution to prevent the algorithm from falling into the local optima.Secondly,the IDPJS algorithm is validated on CEC2017 benchmark functions,and it is compared with five heuristic algorithms.The experimental results show that the proposed algorithm has high precision and good stability.Finally,the IDPJS algorithm is applied to the maximum entropy multi-threshold image segmentation.Image segmentation tests are carried out at threshold levels of 5,7,and 9,respectively.The results show that the proposed algorithm is an effective method to solve the multi-threshold image segmentation problem.（2）Threshold segmentation based on minimum cross entropy has become a common image segmentation technique since it does not require a priori information about the grey of the images.It can effectively perform single threshold segmentation of images,but is computationally expensive when performing multi-threshold segmentation of images.Therefore,this paper proposes a minimum cross entropy multi-threshold image segmentation method based on the ameliorated jellyfish search（AJS）algorithm to improve the computational efficiency of minimum cross entropy multi-threshold image segmentation.In the AJS algorithm,the foraging strategy guides individuals to learn randomly,which speeds the convergence of the algorithm and increases the possibility of jumping out of the local optimum.The dimensional learning strategy constructs a neighborhood for each jellyfish,which makes the jellyfish information shared among the neighbors,maintains the population diversity of jellyfish and effectively balances the exploration and exploitation of JS algorithm.The experiments are conducted on two benchmark functions,CEC2017 and CEC2020,respectively.The AJS algorithm is also compared with six advanced intelligent algorithms,and the results verify that the proposed algorithm is able to obtain high quality solutions.The AJS algorithm is combined with the minimum cross entropy thresholding segmentation method.It is compared with four algorithms for image segmentation at different threshold levels for experiments.The quality of segmented image is evaluated by objective function value,PSNR,SSIM and FSIM,respectively.The results show that the proposed algorithm can obtain more accurate segmentation thresholds and the better quality of segmentation images.