Image registration using particle filter with combined normalized mutual information and correlation coefficient

Image registration has very wide application in various fields of image processing. For comparative study purpose we calculate the mismatch in data using some geometric transform. Considering the application of medical image processing where the image data is obtained from multimodal sources. Many factors are responsible for noise factor. These factors create non-linearity and non-Gaussian nature in noise. Thus, it increases the complexity of problem solving methods. For such difficult problems Particle filter gives better, optimized solutions. Particle filter is a Sequential Monte Carlo method based on Bayesian techniques. It is a global optimization technique. When the optimization using local minima fail give appropriate results global optimization considers all possible samples and compare the results to give better and optimized solution. Particle filter methods are dependent on weight function which is known as importance density function. Here mutual information combined with correlation coefficient is used as the weight function.