Image Fusion And Evaluation Based On Wavelet Transform

Image fusion is an important part of multi-sensor data fusion. It aims at getting more exact and more comprehensive image description of the same scene or object by distilling and integrating the information of multi-source images. Image fusion technology has been widely applied to remote sensing, medicine, computer vision, military and so on.From the abstract level, image fusion can be divided into: pixels-level fusion, features-level fusion and decisions-level fusion. The major research of this paper is pixels-level fusion. The emphasis is image fusion based wavelet transform. At first, the paper introduce the background, the development conditions of image fusion and its widely applications. And then the theory of wavelet transform is introduced. The paper deeply studies fast realization of discrete wavelet, namely Mallat arithmetic. Orthogonal wavelet decomposition of image and biorthogonal wavelet theory are studied. Wavelet transform is used in the image fusion. Some methods, for example simple method of image fusion, principal component analysis, image fusion based on wavelet transform, are compared by subjective and external evaluation.Image is decomposed using wavelet transform, and gets a series of low and high frequency coefficients. High frequency coefficients include detail information of image. Based on average grads reflecting exiguous detail differences of image, the paper proposes a wavelet fused algorithm based on neighborhood average grads. And then the result compared with others two methods such as maximum absolute value and neighborhood variance. Experiments testify that the paper's algorithm is prior to others methods.