A new image feature point detection method based on Log-Gabor gradient feature

High spatial resolution remote sensing image automatic registration is one critical technologies in the imagery processing such as image mosaic, change detection and image fusion. The matching methods based on local characteristics, which applied in high-resolution remote sensing image registration, has the advantages of expressing more local image details, more strong anti-interference capability to the noise and the blocking. In the paper, we analyze the popular automatic registration methods and found the algorithm based on Log-Gabor gradient feature detector and descriptor, then finding the matching point pairs. In the search space we use the angle measure and the whole restrict parameters estimating method to match feature points, estimate image transformation model parameters, and finally achieve high-precision automatic registration results. Through data experiments, it shows that the methods is stable, reliable and more efficient than the traditional algorithms.