Image Fusion Based On The Second-generation Curvelet Transform

Image fusion is a branch of information fusion. Such process generates a single image which contains a more accurate description of the scene than any of the individual source images. This fused image should be more useful for human visual or machine perception, and it could be more accurate and more comprehensive description of the certain image scene through the processing of redundant information and complementary information of the source image. In recent years, image fusion has become a very important research field of many subjects and has got widely application.According to hierarchical classification, image fusion falls into the following three categories:pixel level fusion, feature level fusion, and decision level fusion. The first fusion is the basis of all fusion methods. In this thesis, pixel level fusion is the object of study.According to the image fusion techniques’background and current situation, the thesis gives some basic issues on the image fusion, and discuss image fusion based on second-generation Curvelet. The main work is as follows:(1) Give a brief description of the image fusion techniques in range of applications, current situation, development prospects and difficulties, then points out its significance and importance. On this basis, image enhancement in pre-processing stage based on wavelet transform are introduced and give some simulation results.(2) Image registration algorithm based on mutual information is presented. Interpolation algorithm and Powell algorithm are introduced. At last, experimental result is given.(3) Discuss the fusion steps and comparative study of common image fusion methods. Subjective and objective quality evaluation criteria of image fusion are introduced. Introduce Common evaluation indexes. And give a brief explanation on selecting evaluation indexes. Study the advantages and disadvantages of theory and its applications in image fusion based on wavelet transform, Radon transform, Ridgelet transform, first-generation Curvelet transform and second-generation Curvelet transform. Besides, two methods of the implementations of the second generation Curvelet transform are introduced in detail.(4) Put forward a new image fusion method based on the second generation Curvelet and feature product. Firstly, the fast discrete Curvelet transform is performed on the original images to obtain the coefficients at the corresponding scale. Then the corresponding subband images by using different rules are fused to get the Curvelet coefficient. Fusion strategy includes weighted average in low frequency; adaptively weighted the coefficients in high frequency, while the weight is calculated by feature product which is consisted by local energy, average gradient and standard deviation. At last, the resulting image is obtained by inverse Curvelet transform. The new method is tested by mutli-focus images and multi-spectral images. The comentropy, average gradient and standard deviation are used to evaluate the experimental results. And comparisons with the results based on wavelet transform are also carried out. The test results show that the resulting images based on the second generation Curvelet transform are more clear and better. It can be applied widely.