Font Size: a A A

The New Intelligent Optimization Algorithm And Its Application In Image Segmentation Applied Research

Posted on:2012-01-01Degree:MasterType:Thesis
Country:ChinaCandidate:J H LiangFull Text:PDF
GTID:2208330335971174Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
In recent years, a few of new optimization algorithms based on swarm intelligent are proposed with the development of artificial intelligence and artificial life, in which representative models include Artificial Fish Swarm Algorithm (AFSA), Artificial Bee Colony (ABC) algorithm and Bacterial Foraging Algorithm (BFA). Up to now, these algorithms have not yet been widely used in image processing, because their history is relatively short.Image segmentation is a key step from image processing to image analysis, whose quality will have a direct influence on the consequent image analysis and image understanding. Therefore, the research on fast and efficient segmentation method is always a hot topic for researchers at home and abroad. Emphasizing on the swarm intelligence based optimization algorithms presented after 2002, we analyze their principles and characteristics, and try to introduce them to image segmentation. Our main creative results are summarized as follows:(1) After the behavior patterns of artificial fish, i.e. "Swarm", "Follow" and "Prey" are deeply analyzed, AFSA is improved by adopting the elitism scheme. On this basis, a SAR image threshold segmentation method based on improved AFSA is proposed. In the method, an original image is preprocessed by multilevel stationary wavelet transform firstly. And then, the trace of the between-class scatter matrix is taken as the fitness function of AFSA which is constructed by two-dimensional histogram of the reconstructed image and its mean image. Finally, the best threshold is found out by AFSA. Experimental results indicate that the proposed method has obvious improvement on segmentation quality and segmentation speed.(2) After the behavior patterns of bacterial swarm, i.e. "chemotaxis", "reproduction" and " elimination and dispersal " are deeply analyzed, BFA is improved by shrinking the search space. On this basis, a new SAR image threshold segmentation method based on improved BFA is proposed. In this method, an improved two-dimensional grey entropy model is taken as the fitness function of the improved BFA. The best threshold is located by three behaviors of bacterial swarm, i.e., chemotaxis, reproduction and elimination and dispersal. Experimental results show that the proposed method is superior to some segmentation methods based on Genetic Algorithm (GA) and AFSA in convergence, stability and segmentation effects.(3) Based on the behavior patterns of employed bees, scouts and onlookers in artificial bee colony, a fast SAR image segmentation method based on 2-D grey entropy is proposed, which makes full use of wavelet transform, grey entropy model and ABC algorithm. In this method, a filtered image is produced by performing a second noise suppression to an approximation image reconstructed by low-frequency coefficients in wavelet domain first. Secondly, a gradient image is produced by reconstructing high-frequency coefficients in wavelet domain. Hence, a co-occurrence matrix based on the filtered image and gradient image is constructed. Next a gray entropy model is improved to act as the fitness function of ABC algorithm. Finally, swarm intelligence of bee colony is used to locate the best threshold quickly. Experimental results indicate that the method is superior to GA or AFSA based methods.(4) An image segmentation method based on grey morphology and ABC algorithm is proposed. In this method, grey morphology theory is applied to reduce image noise firstly. Secondly, the fitness function of ABC algorithm is designed by two-dimensional Otsu method. Finally, the best threshold is approached through the parallel optimization ability of ABC algorithm. Experimental results indicate that when the method is applied to segment infrared image or SAR image, it is suitable for the consequent analysis and processing with the ability to provide with good segmented images.
Keywords/Search Tags:image segmentation, swarm intelligence, artificial fish swarm algorithm, bacterial forging algorithm, artificial bee colony algorithm
PDF Full Text Request
Related items