Nonrigid image registration using spatial and multiresolution techniques

In this paper a new approach for nonrigid registration is introduced. The global change in the image is modeled by an affine transformation. Curve let transform is also combined with affine for computation time reduction. Registration error can be obtained between the reference and target image using intensity based similarity measures such as difference, mean square error, gradient difference, log difference, pattern intensity. Registration error for different methods such as affine, affine after Curvelet and Curvelet after affine are done and the results are tabulated.