Non-linear image registration using perspective invariant and thin-plate spine

We introduce a non-iterative geometric-based method to register an image using a novel set of geometric landmarks residing on an extracted 2D contours from the image. These landmarks are intrinsic and computed from the differential geometry of the curve. We exploit the invariant properties of maximum-curvature points that are local and robust under the affine and some perspective transformation. Geometric invariant exploits five-point coplanar invariant constructed from 5 consecutive landmarks. This invariant is preserved not only in affine map but weak perspective map as well. The alignment is achieved by establishing correspondences between the landmarks after a conformal sorting based on derived absolute invariant and undoing non-linear transformation which is a thin-plate spline (TPS). The experiments have shown that the purposed methods are robust and promising even in the presence of noise.