Research on Visual Features-Based Method for Multi-Sensor Image Registration

Multi-sensor image registration is very important in the space image proceedings for its huge applying value. However, there are some difficulties from anisotropic multi-source spatial data. To consider the multi-resolution visual features in the image registration, we take the multi-resolution decomposition based on wavelet to direct the registration from the rough to the subtle, and then collect the contour which could meet the vision characteristics by use of extension of contour tracking algorithm to realize the multi-resolution shape feature matching based on Fourier transform on the basis of outline of the chain code of curvature functions. To compare with the existing image registration algorithm based on characteristics, the results shows that this method has better registration effect to heterogeneous images from multi-sensors