A Region Dense Matching Algorithm for Remote Sensing Images of Satellite Based on SIFT

In the acquisition of 3D terrain information based on images of satellite, image matching is one of the crucial issues. Feature based matching can provide robust results; however, the results are always sparse, and cannot satisfy the need of application. To solve this problem, a region dense matching algorithm for remote sensing images of satellite based on SIFT (Scale Invariant Feature Transform) is presented. In the algorithm, robust matching results of SIFT are taken as the basis, and then matching growing is conducted in different regions of satellite image with affine transformation and least square method. Experiment results show that the algorithm can achieve dense matching for remote sensing images of satellite, especially in few textures or no textures regions.