Research On Applications Of Particle Swarm Optimization In Image Processing

Particle Swarm Optimization (PSO) is a heuristic, optimal search algorithm based on biological principles which has appeared in recent years. It has been widely used in range of industries, because of its feasibility and effectiveness. However, some limitations in the algorithm have been found by many researchers. For the traditional PSO algorithm is easy to fall into local extreme points, many scholars improve it to make search more accuracy and effective, so that it can be used in more areas. One of the simplest and the most efficient improvements is chaotic particle swarm optimization. The basic idea is to transform the variable from the chaotic space to solution space, then use the properties, ergodicity, randomness and regularity of chaos to research. Chaos optimization algorithm search quickens the evolution process, and improves the abilities of seeking the global excellent result and convergence speed and accuracy. This thesis will compare PSO and chaotic particle swarm algorithm through experiment.Image segmentation and image classification are important and basic steps in object examination and recognition. The main objective of image segmentation is to extract objects of interest from an acquired image, which is the fundamental work for image analysis and understanding. In the experiment, we use a two-dimensional Fisher thresholding method. Two-dimensional Fisher thresholding method considers both gray information and spatial neighbor information between pixels in the image, so it is more effective and reasonable. We may regard texture as what constitutes a macroscopic region. Traditional texture classification methods need to extract much texture feature and correct classification is hard to be guaranteed. This thesis will use the Tuned mask to classify images. The main difficulty of the image segmentation and classification algorithms is the high time consuming, which has become an obstacle of the classification development. Therefore, it is important to pursue an efficient algorithm to solve the problem of classification of image segmentation.The main work of this thesis is as follow. (1) The particle swarm algorithm and chaotic particle swarm optimization algorithm have been employed to search optimum two-dimensional thresholds based on Fisher thresholding method. Compared with the exhaustive method, experimental results prove that the traditional PSO and chaotic particle swarm algorithm have greatly improved the efficiency of Fisher thresholding method as the same these two methods can get the best two-dimensional thresholds as that of the exhaustive method. (2) This thesis uses PSO and chaotic particle swarm optimization algorithm to optimize the template for Tuned template. The proposed CPSO enhance the convergence speed and global search capability greatly on different benchmark optimization functions. The algorithm can efficiently obtain the optimal solution through the experiments.