Application Of Improved Bird Group Algorithm In Image Segmentation

In recent years,intelligent algorithms such as genetic algorithm,particle swarm optimization,simulated annealing,and fish swarm algorithm have been applying to solve the optimal values in various fields,and good results can be obtained.The search capabilities of these algorithms are generally limited by the initial population size and various initial parameters.Usually the larger population,the better its search ability,but it takes more time to solve.In this paper,we use the intelligent algorithm of small initial population to solve the image of the otsu threshold.Firstly,the threshold is solved by used six algorithms.By compare the experimental results,we can know that the bird group algorithm the performance in search for otsu threshold is better than the other five algorithms.When image segmented using the otsu threshold,the noise will cause severe damage to the segmentation result.In order to eliminate the influence of different noises,several filters are used to denoising.According to the median-average filter,another positioning-median-Gaussian mean filter is proposed,and the results of the denoising experiment show that the positioning-median-Gaussian filter is superior to other filters in filtering the unknown noise defined herein.For low-density Salt and pepper noise and medium-density salt and pepper noise,the adaptive median filter has better denoising results,but the results of high-density salt and pepper noise is not satisfactory.Fast and efficient mean filter algorithm can perform good denoising on high-density salt and pepper noise,and is improved on the basis of this filter,which enhances the denoising ability of low-density salt and pepper noise.In order to improve the accuracy and stability of the bird group algorithm in solving.In this paper,two improved methods are proposed,one is the genetic variation bird group algorithm which adds the mutation link in the genetic algorithm,the other is the double bird group chaos optimization algorithm which uses the double bird group and adds chaotic perturbation.The improved algorithm is used to solve the one-dimensional and two-dimensional otsu thresholds of several images.The experimental results show that the improved bird group algorithm can achieve the correct one-dimensional threshold within the specified number of iterations.Accurate search of the optimal threshold,the accuracy rate is almost 100%,the result is better than the unimproved bird group algorithm.When seeking the two-dimensional threshold,the solution obtained is closer to the optimal solution than before the improvement,and its stability is also enhanced.Then,using theimproved bird group algorithm and the improved filter,the noise-added image is threshold-solved and segmented.It is found that the segmentation result has higher accuracy.