SAR Image Registration Using Phase Congruency and Nonlinear Diffusion-Based SIFT

SAR Image Registration Using Phase Congruency and Nonlinear Diffusion-Based SIFT

Abstract— SAR Image Registration Using Phase Congruency and Nonlinear Diffusion-Based SIFT. The scale-invariant feature transform < Final Year Projects 2016 > SIFT algo-rithm has been widely applied to optical image registration. ow-ever, mostly because of multiplicative speckle noise, SIFT has a imited performance when directly applied to synthetic aperture adar (SAR) image. In this letter, a novel SAR image registration ethod is proposed, which is based on the combination of SIFT, onlinear diffusion, and phase congruency. In our proposed algo-rithm, the multiscale representation of a SAR image is generated y nonlinear diffusion, since it better preserves edges in the image s opposed to Gaussian smoothing, which is used in the original IFT. To reduce the influence of multiplicative speckle noise, the atio of exponential weighted average operator is used to compute he gradient information in the construction of nonlinear diffusion scale space.Moreover, phase congruency information is utilized to remove the erroneous keypoints within the initial keypoints. Experimental results on multipolarization, multiband, and multitemporal SAR images indicate that our algorithm can improve the match performance compared to the SIFT-based method, which leads to a subpixel accuracy for all the tested image pairs.