Multi-Resolution Remote Sensing Image Registration Based on Tensor Voting

Abstract

Edge feature is one kind of very important feature in multi-source remote sensing image registration and is the foundation of image registration based on edges. An algorithm is proposed, which makes full use of tensor voting and multiresolution algorithm to extract the large edges of remote sensing images, and combines the accurate localization of the tensor voting in high resolution and the anti-noise performance in low resolution after wavelet transformation to detect more complete edges. The proposed method has provided one kind of steady match characteristic for the multiple source remote sensing image registration.