Research Of The Wavelet Transform Image Fusion Algorithm Based On Image Sequences

At present, domestic and foreign have obtained great achievements in image fusion research. Image fusion is that combine two or more than two sensors on a particular scene from multi source image information to be extracted and integrated,so that it can get the image information of single sensor which is not available. The aim is to reduce uncertainty. Wavelet transform as a kind of image fusion methods,and it’s characteristics of time domain and frequency domain localization are very good, so it has great application value of wavelet transform in image fusion, and it is the hot spot in the research of image fusion. The main work in this paper is the image of the wavelet transform image fusion algorithm based on exploration,it tells the story of some traditional fusion methods, and puts forward the corresponding improvement algorithm, using the methods of combining theoretical analysis and simulation experiments, some meaningful results are obtained. The specific contents are as follows:(1)Consult a lot about image fusion methods, describes the current research status of image fusion.(2)Introduce the basic knowledge of image fusion, mainly introduce the relevant theoretical knowledge of image fusion principle, classification and wavelet transform fusion.(3)Systematically describe the pretreatment work before image fusion, especially in image denoising and matching are described in detail. For image denoising,this paper puts forward a kind of lifting wavelet transform denoising method, wavelet threshold method was used to quantify, the principle is the combination of lifting wavelet fast calculation speed, and memory is small, and the image denoising. For image registration, we present a method of lifting wavelet transform characteristics,the principle is that do the preprocessing of the original image, and then using the wavelet transform modulus maxima edge detection method for image edge detection based on edge feature point extraction; Finally,use normalized cross-correlation matching and early least squares matching accurate registration of image.(4)Study the conventional image fusion algorithm, and based on the simulation experiments, by comparing the image after image fusion processing, points out the shortcomings, in view of this,this paper puts forward a kind of improved wavelet transform fusion algorithm. The improved method is as followed: firstly, wedecompose the source image wavelet and get the low and high frequency compoents.Secondly, we deal with those two components by using different fusion rules. During this procedure, the low frequency part is dealed by using weighted coefficient method based on the local energy while the high one the regional variance method.(5)Wavelet transform plays an important role in image sequences fusion methods,but its slow operation speed, complexity, in real-time applications, especially in the time space,it will get limit. Therefore, this article on the former basis wavelet algorithm, an improved algorithm is proposed based on lifting wavelet transform, the principle is that the low frequency components after wavelet decomposition by fusion method combined with the weighted average, using the fusion strategy of combining variance of the wavelet coefficients and the absolute value to deal with the high frequency components. In order to further illustrate the proposed fusion algorithm,through the experimental comparison with traditional fusion methods, the subjective analysis and objective performance evaluation together carries on the detailed analysis of the experimental results, so that the algorithm can be practical, efficient and superior.