A Multiscale Adaptive Rift Method for Low Coherence Sar Image Matching.

In this paper, an improved Radiation-Variation Insensitive Feature Transform (RIFT) method is proposed for low coherence SAR image matching. First, pyramid scale space is introduced in the period of feature transformation, which can extract features with different scales, increase the number of feature points and achieve the scale invariance of the algorithm. Then, the key parameters of the RIFT algorithm are calculated and selected adaptively according to the sizes and statistical characteristics of the images, so as to acquire feature points with even distribution. Finally, we conduct experiments on SAR images obtained at different wavebands, platforms and times. The experimental results show that, the proposed method can obtain more feature points with more even distribution, and the matching accuracy is also improved.