Hybrid Particle Swarm Optimization Algorithm In Image Fusion

Image fusion is one of the important parts of information fusion, the image from Different image sensor and the same sensor can obtain redundant and complementary information. Image fusion combines all those different information to acquire an new image that satisfied with some special application. It aims at getting more exact and more comprehensive image description of the same scene or object by integrating the information of the multi-source images. Image fusion is widely applied in a variety of fields such as remote sensing, medical image analysis, automatic target recognition, intelligent robots, military affairs, computer vision and so on. So image fusion is a useful technology. Image fusion methods have its thresholds value and parameters, different values have different fusion performances in common. But the thresholds value and parameters of the fusion method are usually configured to rely on the subjective experience, which have significant influences in fusion algorithm performance. It has profound practical significance of optimizing the thresholds and parameters.Particle swarm optimization (PSO) is an optimization algorithm based on swarm intelligence, which is the same to other evolutionary computations and developed by Kennedy and Eberhart in 1995. Particle swarm optimization has become the hotspot of international evolutionary computation because of its simple concept, easy for implement and excellent performance. First, the paper deeply researches the PSO, introduces its principle, basic process, settings of parameters, using process and so on. Aiming at the disadvantage of its fast convergence in early, low precision in late and so on, this thesis analyzes several improved algorithms and briefly describes the application of some PSO algorithms in image fusion recently in introduction. Next, the paper reaches image fusion, analyzes its typical algorithms, subjective and objective evaluation criteria and fusion rules.Aiming at the question of image fusion, the thesis presented a image fusion method based on wavelet transform and block segment of the images. An immune particle swarm optimization search algorithm was developed for fusion of two spatially registered images. In order to get the optimal images, the size of block was defined as particle. Immune particle swarm involves the immune information processing mechanism of immune system into particle swarm optimization, which can improve the abilities of seeking the global excellent result. A self-organizing particle swarm optimization search algorithm was applied in image fusion method based on region energy. In order to get the optimal images, the size of threshold value was defined as particle. Self-organizing particle swarm optimization comprehensively considers acceleration coefficients and inertia weight which are two important parameters in particle swarm optimization. According to adaptively adjusting acceleration coefficients and inertia weight, the particles are organized to track the domain of attraction of local optimum and the domain of attraction global optimum respectively during the search. Self-organizing particle swarm optimization is developed for solving premature convergence of particle swarm optimization. The performance of the methods proposed in this thesis was better than the performance of the existed fusion methods, which has important significance in researching and developing image fusion. W-PSO was tested for performance comparison with IPSO and SOPSO.