Multilevel Image Segmentation Based On Improved Bat Algorithm

Image Segmentation is an important part of image processing,and the quality of segmentation results will affect the accuracy of image analysis.At present,there are two main problems in image segmentation.First,there is few methods to adapt to image segmentation under most conditions.Second,there is no unified standard for evaluating segmentation results.All the problems make the research of image segmentation important.Segmentation based on thresholding is the most widely used in image segmentation.More and more researchers combine the swarm intelligence optimization algorithms(particle swarm optimization ? differential evolution algorithm,etc)with the traditional threshold segmentation method.The solution corresponding to the extreme value of the objective function is taken as the optimal threshold for image multilevel image segmentation.Bionic algorithms,including Cuckoo Search(CS)and Firefly Algorithm(FA)are emerging algorithms in recent years.Different from the traditional intelligent algorithms,they can achieve better performance with fewer parameters and better mechanisms.They are more accurate and faster in solving complex problem optimization.Combining bionic algorithms with traditional threshold segmentation algorithms can effectively reduce noise pollution and time loss,improve operational efficiency.This paper studies the bionic algorithms and applies them to multilevel image thresholding.The main work is:(1)This paper introduces the principles and mechanisms of three bionic algorithms,and verifies the optimization accuracy and operational efficiency of the algorithms by simple experiments.The improved three algorithms are used to find the optimal values of 21 benchmark functions.The optimization performance(accuracy,time complexity,etc)of the three improved algorithms are compared and then select the bat algorithm,which has a better performance for in-depth research.(2)The frequency in the bat algorithm is randomly generated.For the frequency shift phenomenon in real life,considering the problem that the algorithm is sensitive to the initial parameters,the frequency shift effect is introduced into the bat algorithm to improve the frequency,effectively reducing the optimization time and improving accuracy.The simulated annealing weight is introduced into the global search to increase its global search capability.(3)Combining the improved bat algorithm with the maximum inter-class variance and maximum entropy,multilevel image thresholding based on improved bat algorithm is proposed.The improved bat algorithm is used to obtain the optimal segmentation threshold of the image,and then compare the results with the particle swarm algorithm and bat algorithm.Test and verify the effectiveness of the improved algorithm results by image quality evaluation criteria.