Image Fusion Based On Wavelet Transform

In recent years, image fusion has become a important and valuable and new technology of Image Understanding and Computer Vision fields . The main aim of the image fusion technology is to integrated image information which obtained by different types sensors ,and improve the reliability and the observability of the image by processing the redundant datas among the images .We can raise the result of image by processing the complementary information. Image Fusion technology has been widely applied to geology, agriculture, mapping, computer vision, military, remote sensing, medicine, biology and so on. With image fusion in-depth technical studies ,it will be broader applications.From the abstract level, image fusion can be divided into : pixels-level fusion, feature-level fusion and decision-level fusion . The major research of this paper is pixels-level Image Fusion based Wavelet Transform. Because Wavelet Transform have the characteristic of multi-resolution or multi-scale . We can divided images into a series of frequency space by wavelet decomposition .and then process the image information in some fusion rule in every frequency space .We can get the fusion image through Wavelet rescontruction in the end .The image fusion based on PCA transform also be studied . The fusion image based on wavelet transform can be achieved better results and faster processing speed.This paper have studied about the Performance Evaluation of fusion image, and establish the performance scheme based on information and statistical characteristics and signal to noise ratio and gradient value .It compared multiple fusion scheme on the use of these Performance Evaluation scheme . What is more ,this paper have learned and studied the theory and practice about image registration ,and establish the similarity measure based on Mutual information .Image get best registration when it get the maximum Mutual information . A method of registration paremeters's optimization search based on wavelet transform is proposed .It's proved that image registration base on mutual information could be achieved a good result and quick-speed if it use the optimization algorithm base on wavelet transform .