Application SFLA Multi-threshold Image Segmentation

Shuffled Frog Leaping (SFL) algorithm is one of the newest optimization algorithms of swarm intelligence presented in the21st century. Since the algorithm adopts a search strategy which combines the global search with local search, it has an efficient ability to search in parallel. However, as a newly-presented algorithm, its application on image processing is seldom reported. Taking the SFL algorithm as the research object, this dissertation analyses the search mechanism and the basic principle of the SFL algorithm, and then introduces it to the image multilevel thresholding segmentation successfully. The creative fruits mainly focus on the following issues:(1) Using the efficient ability of the SFL algorithm to search in parallel, an image multilevel thresholding segmentation method is proposed. In this method, the Otsu method is employed to serve as the fitness function of the SFL algorithm in evaluating the quality of the segmentation threshold represented by each frog, and guides the frog group in moving towards the optimal solution. Experimental results show that the proposed method can not only provide good segmented images, but also is superior to some segmentation method based on Artificial Fish Swarm (AFS)algorithm in terms of segmentation quality and segmentation speed, especially suitable for searching more thresholds.(2) An improved SFL algorithm is put forward based on multi-individual fuzzy perception (MFP shuffled frog leaping algorithm), on which a multilevel thresholding segmentation method is further proposed. The proposed segmentation method not only expands the reference information of the individual movement, but also introduces fuzzy control to update the individual so that the algorithm can dynamically adjust the jump step of the current individual with iteration times, which ensures the flexibility of the algorithm. Meanwhile, the operation of updating the worst frog of each community continuously is changed to update all the frogs of each community, which can improve the efficiency of the population migration. The results of our experiments on some medical images show that when taking the Otsu method as the fitness function, the MFP shuffled frog leaping algorithm is better than some other swarm intelligence based algorithms, such as basic SFL, PSO, ABC and AFS, in terms of optimal thresholds, segmentation quality, segmentation time and stability of algorithm, especially suitable for searching more thresholds.(3) An improved shuffled frog leaping algorithm is developed based on immune evolutionary and boundary mutation (IEBM shuffled frog leaping algorithm), on which a multilevel thresholding segmentation method based is suggested. Refering to the experience of the immune evolution theory, the optimal frog individual is evoluted in each iteration in the IEBM shuffled frog leaping algorithm, which is of advantage to the production of the optimal solution. The idea of boundary mutation not only guarantees the individual in the search area, but also avoids the solution gathering at the boundary and falling into some local optimum. In addition, the idea of fuzzy control is used to dynamically adjust the jump step of the frog. Taking the Otsu method and the Maximum entropy method (Kapur method) respectively as the fitness function of the IEBM shuffled frog algorithm, experimental results indicate that the proposed method has more advantages than some other swarm intelligence based algorithms like basic SFL, PSO, ABC and AFS, especially suitable for searching more thresholds.