Multi-Threshold Image Segmentation Based On BBO Algorithm

The image segmentation is a key step in image information mining.Threshold segmentation is popular because of its simple and effective computation.The one dimensional multi-threshold segmentation algorithm does not take into account the local information of the image.When the image scene is more complex,it is hard to get a better segmentation result.The computational complexity of two-dimensional multi-threshold segmentation algorithm is relatively high,and finding the optimal threshold combination is difficult.When exhaustive method is used for traverse search of the two-dimensional multi-threshold segmentation algorithm,it leads to high space and time complexity of the algorithm.The introduction of intelligent optimization algorithms is a good solution.The biogeography-based optimization algorithm is a relatively new intelligent optimization algorithm developed in recent years.Because BBO algorithm involves few parameters and simple computation,plus its unique search mechanism,it has attracted the attention of many scholars.Although BBO algorithm has many advantages,it also has its limitations.Because the traditional migration operation easily leads to the appearance of similar individuals in the population and weakens the ability of the algorithm to search the optimal solution,the algorithm is prone to premature convergence in the later search and obtains the local optimal solution.In addition,the algorithm performs mutation operation in a random way,which limits the global search ability of the algorithm to some extent.In order to solve the problem of high space and time complexity in multi-threshold image segmentation,in this thesis,the BBO algorithm is introduced to optimize the problem.Besides,in order to improve the global search ability of BBO algorithm in multi-threshold image segmentation,an improved BBO algorithm is proposed.The main research work is as follows:(1)In this thesis,the thresholding method of the maximum entropy and exponential cross entropy for image segmentation is studied.The two-dimensional exponential cross entropy threshold segmentation criterion in condition of oblique state is described in detail,and the oblique division method is extended to multi-threshold segmentation.(2)In order to improve the global search ability of BBO algorithm,based on its advantages and disadvantages,an improved BBO algorithm is given.First,the elitist selection operator is applied to the BBO algorithm,and a few sets of optimal solutions are retained.Secondly,a migration strategy based on the selection operator's optimal solution and the random solution fusion method is proposed to reduce premature convergence and invalid migration of traditional migration operation.Finally,binary mutation operation is presented to reduce blindness of traditional mutation operation.(3)The improved BBO algorithm is applied to two-dimensional exponential cross entropy multi-threshold image segmentation.In order to reduce the computational complexity of the two-dimensional exponential cross entropy multi-threshold segmentation,the oblique method is used in this thesis instead of using the traditional direct method.(4)On the platform of MATLAB,simulation experiments of two-dimensional exponential cross entropy multi-threshold image segmentation based on the improved BBO algorithm are carried out.The algorithm is compared with other intelligent optimization algorithms,and the results show that the algorithm has the better segmentation effect.