Medical Image Registration Based On Mutual Information Technology Research

With the development of science and technology, the way of accessing to image date was also ever-changing. When people took advantage of these image data, a single image data often can not meet the requirements of the people in the practical application. These image data was reasonable used as much as possible that made image integration results were more accurate. It made the integration accuracy and image recognition efficiency rapid increase. Image integration technology was applied in the image processing from the multiple sensors, Image guidelines was a specifically for the image data processing technology. This dissertation studied the registration techniques of mutual information-based medical image.A detailed overview of image registration techniques basic process and the basic classification. This dissertation especially in-depth studied the technology principle of the algorithm, the algorithm steps and algorithms to achieve. This dissertation also optimized and improved the main algorithm technology.(1) In this dissertation improved a threshold denoising algorithm based on multiwavelet. Based on the original algorithm selection of a threshold, this dissertation used of two thresholds. In the dissertation defined a new differential slope threshold function and successfully applied to improve the original algorithm based on wavelet threshold denoising algorithm. Finally, the experimental simulation confirmed this dissertation improved algorithm for image denoising is better.(2) After the removal of image noise information, the image outline used rough registration in this dissertation. Get approximate search range of image optimization algorithm from rough registration. Finally, register image based on mutual information. Thus it is critical to the successful extraction figure want the edge contour. In the dissertation, the image edge detection method based on wavelet and classic edge detection method such as based on the effect of the Roberts operator, Sobel operator, Prewitt operator, Laplace operator, Canny operator of edge detection and edge detection method compared the effect of edge detection. Experimental results showed that the wavelet-based edge detection algorithm had a better edge detection effect.(3) This dissertation also proposed an improved genetic algorithm and improved optimal Preservation strategies. First analysis of the relevant principles of the basic genetic algorithm and the main steps of the algorithm, Secondly, according to the inadequacies of the basic genetic algorithm, gave a better adaptive genetic algorithm. The improved algorithm search performance was improved. Mainly through the adaptive adjustment of individual cross and the mutation probability to optimized and update. In addition, using the improved elitist strategy not only made the search for the optimal solution more closer to the global optimum but also made it difficult to fall into local extreme points. Found after the value of each pixel in the image after a series of operations. According to bilinear interpolation algorithm that interpolation results were good, final output image after registration completed(4) This dissertation gave program design of the registration process. It selected the represent the meaning of medical images to test and compared the integration effect with other image registration method.