Study Of Image Fusion Based On Contourlet Transform

Image fusion is a process that could fused two or more images into one image according to a certain algorithm, which can produces more abundant,precise and reliable information than a single image. The object or scene could be completely described by data information that is produced from image fusion. Contourlet transform is a framwork to calculate the multiple directions and multiscale of discrete images, which is proposed to apply in the two dimension image processing. Compared with wavelet transform, Contourlet transform has the ability of not only local time-frequency analysis but also stronger directional identification, which can represent the two dimensional image with higher approximation precision and better sparse expression. The features of original images could be extracted more closely in image fusion by Contourlet decomposition. The applications of image fusion based on contourlet transform is investigated in this thesis. The main results in the thesis are as follows:Firstly,the principles of contourlet transform are presented detailed. The applications of contourlet transform are discussed and a new image fusion algorithm based on contourlet transform and regional similarity is proposed. Multiscale decompositions of source images are carried out with contourlet transform at first. Then the low-frequency coefficients and high-frequency coefficients are fused based on the rules of regional similarity separately. At last, the fused coefficients are reconstructed by contourlet transform to obtain the fused image.Then a algorithm of region-based image fusion using blocks segmentation is proposed which is applied to the multi-focus images. The high-frequency coefficients are fused based on the algorithm of region-based. The low-frequency coefficients are fused based on the algorithm of arithmetic mean.The experiment results show that the fused image effects based on the algorithm of contourlet transform are better than based on the algorithm of wavelet transform. Compared with the other methods,the two algorithms in this paper could obtain better fused effects.