Improved PSO Algorithm In Image Registration

Swarm intelligence from the last century since its birth, it has been tremendous development in the field of the optimization. Some scholars studied the behavior of biological communities and derived a lot of inspiration, which proposed optimization algorithm based on swarm intelligence, one of which was the particle swarm optimization (PSO, Particle Swarm Optimization), and is global optimization algorithm of a parallel genetic algorithm. This algorithm is a heuristic search strategy based group’s global optimization algorithm. After the proposed algorithm it has got a lot of attentions, and gradually becomes a hot topic, because the particle swarm optimization (PSO) algorithm is simple and with fewer parameters. Based on these characteristics, this approach has some smart and social. Now, PSO is a wide range of applications in areas such as signal processing, neural networks, pattern recognition, image processing, parameter optimization, data mining, and has achieved good results.Image registration is the current hot spot in the image processing in large part because it is the premise of image fusion and has important research value. The image registration is defined as two or more images, different times, different sensors (imaging device), or under different conditions (weather, illumination, imaging position and angle, etc.) to obtain the match superimposed. The image registration refers to a geometric transformation on the plurality of images in different times, in different scenarios space, thus making represent the same structure as the pixel in space to achieve consistency. In the past few years, the technology has been development rapidly. On the image registration step, which generally includes three steps:First, determine the coordinates of the registered image with the floating image, and define the relationship between the coordinate systems; next, define the similarity measure as the criterion; finally, application optimization algorithm, where in the second and the third step is the core of the image registration? The floating image and the reference image of similarity measure we has selected peak signal-to-noise ratio (PSNR).This paper is summarized as follows:First, it has a brief description of the image registration and the theoretical basis of the standard PSO algorithm and research, image registration and the principle of the PSO algorithm, the algorithm process. Second, the article has improved the standard PSO because the standard PSO easily falls into local optimum. On the one hand, because it easily traps in local optimal problem of the so-called "precocious" phenomenon, we have proposed an evaluation criterion:Particle similarityδ(t).On the other hand, for the loss of particle diversity it can bring the phenomenon of "premature", we has introduced this concept of average populations of most value Pbavg.Finally, the improved PSO algorithm applied to the PSNR to optimize the similarity measure in image registration, and was simulation comparison with the standard PSO algorithm.From the result, we can see the improvement of the superiority of the PSO algorithm to improve the error.