Research On Multi-mode Medical Image Fusion Technology Based On Multi-channel Transformation

The recognition and understanding of images are the cores of medical imaging automatic diagnostic, with great value and significance in the area of medical image researching. Medical images of the human body contain a wealthy image information and rules, so that medical image illustrate, comprehension, application are faced with enormous challenges. How to get a more comprehensive understanding of the same scene image information and how to apply two or more imaging equipment available, are becoming hot topics for future research. This paper, focus on medical image fusion, one of the hottest issues in medical image processing research carries out as follows:(1) According to the way of the human visual system and the characteristics of image understanding, combined with multi-resolution fusion method, the HVS-WB medical image fusion rule is raised as improvements. This rule uses different ways to combine human visual perception habits, takes different mode decomposition into different parts of the fusion method, improves the traditional multi-channel decomposition in the high frequency part and the low frequency part of the fusion rules consistent. Improving the fusion rate, as well as taking into account the precision of the fusion.the improved method is more robust and more accurate than traditional image fusion method.(2) As to pixel-level medical image fusion, proposes an algorithm based on segmentation of significant areas. By considering the impact of target to the whole algorithm, uses significant regional segmentation to the fusion rule improvement, combing with the characteristics of the high-frequency energy information from multiscale decomposition, so that the results are more consistent in fused images, better to meet the needs of image quality and effectively, solving the problem of losing details in the processing of fusion.(3) First time introduces Framelet Transform to medical image fusion, combined with improved image fusion rule in HVS-based image fusion, proposes the image fusion algorithm based on Framelet Transform. By using improved algorithm, fusion image obtains the highest quality, most clear issue of edge contour. Furthermore, details of the source image are preserved maximize. This solves missing edge information and texture fuzzy problem in traditional multi-iteration fusion based on wavelet image fusion process.(4) Propose medical image fusion algorithm based on the combination of Second-generation Curvelet Transformation and pixel contrast energy feature. The algorithm takes advantages of the characterization of the image curve in Curvelet decomposition, combined with the energy features of each pixel in the partial window. Through experiment, analyzes the curve information and the edge portion superiority in fusion image, solveing the wavelet transform can not effectively acquire geometric features from the image.(5) Propose the concept of Multi-iteration medical image fusion. For the different sources image characteristics uses Multi-iteration fusion method by selecting appropriate fusion algorithms. Experiments show that a reasonable choice of the source image used in multi-iteration fusion method, can effectively solve the problem that losing details by a single traditional image fusion method, it can also make up information missing in single image fusion method. The objective indexes of evaluation are all improved.