A Research On The Algorithms Of Registration And Fusion For Multi-spectral SAR Images

Multi-spectral SAR images provide complementary and redundant information. Fusing the information of different images can get more complete and reliable information than only using one single image.In this thesis, we mainly discuss the algorithms of registration and fusion for multi-spectral SAR images, including the edge feature extraction algorithms, the registration algorithms and the fusion algorithms.Edge feature extraction in SAR images is beneficial for the registration of SAR images. However, edge detection, in single look SAR images is rather difficult due to the presence of speckle. An edge detector based on multi-oriented plate and the students' test is presented in this paper, which is proved to have the ability of extracting edges from SAR image with high detection performance and localization ability. Then the algorithm is extended to extract the line features from the SAR images.Image registration is the basis of image fusion. To cope with the different appearance of different spectral SAR images, three algorithms are presented for the registration of SAR images: the improved registration method based on the control points, the FFT based method using edge features; the HT(Hough Transform) and FFT method using edge feature. The three methods are presented to cope with different distortion between multi-spectral SAR images.In the end of the thesis, a fusion method based on multi-wavelet transform for multi-spectral SAR images is discussed. Tests prove that the method is better than most of the present fusion algorithms.