The Research On Multi-source Image Fusion Algorithm Which Based On Wavelet Transform

Image fusion is an important part of image processing and computer vision. It cantake advantage of the same scene image which is obtained from a variety of sensors toget an image which is more suitable for human vision and further processing andanalysis of computer. In recent years, multi-resolution analysis represented by waveletis used in the field of image fusion more widely. Fusion rules based on traditionalwavelet analysis does not take the noise and the edge of the image information impacton the fused image into account, therefore, to a certain extent, it can be cause the blurrededge of fused image, sharpness reduced, some details loss, noise increasing and otherissues.Therefore, the research on image fusion technology is a subject which is ofimportant theory and applicational value.This paper introduces the basic concepts of image fusion theory and waveletanalysis theory, and also the subjective evaluation criteria and objective evaluation ofimage quality in details.In order to solve the issue that the edge information and details easily lost in themulti-focus image fusion process, an improved algorithm for multi-focus image fusionis proposed in this paper. The choice of different fusion rules is decided by differentcharacteristics of the frequency domain. The low-frequency coefficients are given basedon the information of the image edge. The high-frequency coefficients are given basedon the variance of window and the regional information entropy. The experimentalresults demonstrate that the image fused by the new method is better than approaches inthe past both in visual effect and objective evaluation, which can be more effectivelyapplied to multi-focus image fusion.According to the characteristics of infrared light and visible light images, thispaper proposes an improve algorithm based on the rules of regional energy fusion. Thelow-frequency coefficients are given based on image fusion rule area of energy;high-frequency coefficients are used for the integration of the rules window variance with regional energy. The experimental results show that the improved fused image,whether in visual effects or objective evaluation, is better than previous results obtainedby fusion algorithm obtained.