Space image registration algorithm based on nonsubsampled Contourlet transform and MLESAC

To superimpose the space images which are taken at different times, an algorithm based on nonsubsampled Contourlet transform (NSCT) and maximum likelihood estimation sample consensus (MLESAC) for space image registration is proposed. Firstly, in order to extract the edge characteristics of the stars, the space images are transformed by NSCT, and then the feature triangle whose vertexes are the mass of stars is structured according to a certain rule, and the triangle is matched according to the guidelines, and then the centers of gravity of the used congruent triangles are validated by MLESAC. Finally, the matched feature points are brought in the affine transformation model to obtain transformation parameters and the registration image is gained. The method, which reduces the complexity of the classic methods underground remaining the registration accuracy, can avoid the effects of illumination changes and noise. 50 teams' space images are used to be vivificated, and the results show that the algorithm can effectively suppress noise, light and space images circumstances, the RMSE of the registrated images can achieve to 0.3741.