Research On The Improvement Of BBO Algorithm Applied To Multilevel Image Thresholding

Image segmentation is the basic process of digital image processing and computer vision, and is a important key step in image processing. Image thresholding is a simple, stable and widely used segmentation technique which is based on the gray histogram of image. But multilevel thresholding needs to find the best group of thresholds among the whole gray level, this process needs a large amount of calculation, costs high computational time. Therefore, it is very important to optimize the search process of multilevel thresholding by optimization algorithms. Biogeography-based optimization(BBO) algorithm is a new kind of swarm intelligence based evolutionary algorithm, it simulates the rule of the population generation, migration and extinction between habitats. BBO algorithm only needs to set up a few parameters, is easy to calculate and converge fastly. BBO algorithm attracts widely attention in the field of intelligent optimization due to its excellent performance and its unique search mechanism. There is no doubt that BBO algorithm has a lot of advantages, but also has its limitations. BBO algorithm realize information sharing between the candidate solution via a migration operation, but this operation has certain blindness affect the exploitation ability of the algorithm. Besides, random variation based mutation operation limited the exploration ability of the algorithm. As a result, BBO algorithm tends to premature convergence and to trap in a local optimal during the optimization process. To overcome the disadvantages of multilevel thresholding such as high computational complexity and time-consuming, the BBO algorithm is introduced in to optimize the thresholding optimization process. In order to enhance the global search ability and improve the performance of BBO algorithm in image segmentation, and to solve the problems discussed above, the BBO algorithm is studied and improved in this paper. The main research work and innovation of this paper are as follows:1) To improve the performance of BBO algorithm comprehensively, an polyphyletic migration and adaptive mutation based BBO algorithm was proposed: an polyphyletic migration operator was introduced to improve the global search ability, and a new adaptive mutation operator was presented, and a greedy selection operator was used instead of the original elitist selection operator to accelerate the convergence process. Then, the polyphyletic migration and adaptive mutation based BBO algorithm was applied to multilevel thresholding based on Shannon entropy to obtained more accurate thresholds.2) In view of wide range multilevel thresholding, a BBO algorithm with dynamic migration and salt & pepper mutation(DSBBO) was proposed, and applied to 2 to 8 thresholds segmentation based on minimum cross entropy, and achieved an effective multilevel thresholding optimization method.3) In order to reduce the computational complexity of the BBO algorithm and speed up the optimization process, the mutation operator was hybrid into migration operator to simplify the processes of the BBO algorithm, called simple BBO algorithm. This simple BBO algorithm was applied to multilevel thresholding based on OTSU to get a fast multilevel thresholding method.4) The improved BBO algorithms introduced above was integrated and used in the multilevel thresholding image segmentation system, the system was realized based on MATLAB software.