The Research Of Image De-noising And Fusion Algorithms Based On Wavelet Transform

The image is often polluted while collecting and transferring. So, it’s necessary to de-noise before processing and analyzing the image. One image of a scene captured by a single sensor may not contain all useful information that people are concerned about due to the different imaging modalities and environment factors. The technology of image fusion can solve this problem by fusing several images of the same sense. The analysis method based on wavelet transform is a time-frequency analysis method. It has good time-frequency localization ability, and plays a very important role in image processing. In the field of image de-noising, wavelet transform can separate noise and useful information so that the image can be de-noised simply. In the field of image fusion, after decomposing several original images using wavelet transform, low frequency parts reflect the outline of image and high frequency parts reflect details of image, and coefficients in different direction reflect edges in different edges. Using different fusion operators and fusion rules for low frequency parts and high frequency parts can dig up redundant and complementary information so as to acquire good fusion effect.This paper studies image de-noising and fusion algorithms based on wavelet transform, the idiographic work is as follows:(1) Analyze image de-noising based on wavelet threshold. On the basis of summarizing several threshold functions, this paper improves them. Then uses MATLAB to compare the de-noising effect of several existing threshold functions and de-noising effect of the new threshold function proposed in this paper, the results show that the method proposed in this paper does better in de-noising.(2) Analyze image fusion algorithms based on wavelet transform. This paper makes an improvement on the basis of summarizing several existing algorithms. In the field of low frequency, as low frequency parts contain most information of the image, this method chooses local information entropy and local standard deviation as a fusion operator, and makes the consistency checking on the basis of choosing the larger fusion operator. In the field of high frequency, as high frequency parts reflect significant feature such as the edge of the image and coefficients in different directions reflect different edges in different direction, this method defines different significant variables for coefficients in different directions and makes the consistency checking on the basis of choosing the larger significant variable. Finally, this paper uses MATLAB to compare the image fusion algorithm proposed by this paper and several existing algorithms. The result shows that the method proposed in this paper is better in fusion effect indicators such as contrast, information entropy and definition of the fused image.