Multifocus Image Fusion And Evaluation Method Analysis Based On Wavelet

Image fusion is a kind of image processing techniques, which is the fusion of multi-sensor information from vision. The fusion of multi-focus image is a reprehensive subject in image fusion domain. It helps to release from the focus length limit of optical lens in a CCD so as to acquire a clear image of all relative focused objects. The quality of fused image is better than any one of original images. Consequently the utility ratio of image information and the reliability of object recognition are improved. At the same time how to evaluate the effectiveness of various image fusion methods and the quality of fused result so as to make the selection of fusion algorithm performed under these quantitative measures has become an important research project.This paper investigates and discussed the principles of several image fusion methods. Wavelet analysis theory and fusion methods based on wavelet are introduced in detailed. Moreover the number of decomposition layers is selected. In succession, after the introduction of several evaluation methods for image fusion effect, a new evaluation method based on fuzzy logic theory is presented. Entropy, mutual information, average gradient and difference are chose to construct a single factor measure set for the evaluation of multi-focus image fusion. Membership Functions is determined by human knowledge based on expert judge method. An overall evaluation measure is synthesized using several single factor measures by contrast sorting methods.Several fusion methods is compared at the experimental part, wavelet method is proved to have advantage. Compares the results of different decomposition layers, and the best number of decomposition layer is determined to be 4 or 5. Finally evaluation method based on fuzzy logic theory is applied to the experiment of multi-focus image fusion. Result of the experiment and theoretical analysis is consistent with the visual effect valuation of humans, so the validity and stability of this method are proved.