Multi-scale image warping using weighted Voronoi diagram

A new multi-scale image warping method based on the weighted Voronoi diagram is proposed. Weights are assigned to the control points according to their influence scales. At each scale level, a triangulation based on the weighted Voronoi diagram is constructed. Then the interpolation of displacements is performed on this triangulation. In this process, only the control points with large weights are mapped to their final values. Once the control points with large weights have been mapped correctly, their weights are modified to be the maximum weight of those unmapped control points and their displacement values are set to 0. This process is performed iteratively until all control points are correctly mapped. The advantage of this approach is that the underlying triangulation changes between scales to fit the warping scale. Both global warping and local warpings can be modeled appropriately using this approach.