The Research Of Multi-level Fusion Technology About Image Information

The information fusion of the images is integrated with a scene from multiple images, or the same image multi angle information, to be more suitable for human visual perception and computer vision image, or more suitable for further image processing need image. Image fusion system have achieved success in related fields because of its redundant, complementary advantages, time and relative low cost and highly valued,and are able to dig potential in medicine, computer vision, remote sensing, weather forecast, target detection and recognition and other applications.This article mainly aims at the image information fusion at all levels(pixel level, feature level and decision level),launched research in system and depthly. This paper relates to algorithms of face recognition, algorithm of feature level and decision level fusion, system of fusion evaluation, and a new method based on pixel level fusion. The main work of this thesis includes:1.In the pixel level fusion, analysis the reason of the color distortion in the HIS transform, Put forward a combination of methods about the wavelet transform and HIS transform, solute the fusion of color distortion in effective, evaluate the experimental results according to the fusion evaluation. The results show that, this method is a more ideal, promote the research of the pixel level fusion.2.In feature level fusion, using principal component analysis (PCA) method and kernel principal component analysis (KPCA) method to extract feature vectors for face recognition respectively, get the two order statistics and high order statistics of a face image; according to the requirements, constructing a suitable BP neural network fusion the two kinds of feature vector, and compare the fusion vector recognition results with the results before. Validation the advantages of feature level fusion and is suitable for real-time processing, Lay the foundation for the follow-up feature level fusion research.3.In decision level fusion, solve the decision fusion model and problem in effective fusion by using fuzzy set theory. establish fuzzy inference system by using the GUI tool in MATLAB, fusion the image recognition decision effectively by considering the objective fact (face recognition) and the degree of trust (variable membership).