An improved radial basis function based method for image warping

Abstract

A new image warping method is proposed in this letter, which can warp a given image by some manual defined features. Based on the radial basis interpolation function algorithm, the proposed method can transform the original optimized problem into nonsingular linear problem by adding one-order term and affine differentiable condition. This linear system can get the steady unique solution by choosing suitable kernel function. Furthermore, the proposed method demonstrates how to set up the radial basis function in the target image so as to achieve supports to adopt the backward re-sampling technology accordingly which could gain the very slippery warping image. The experimental results shows that the proposed method can implement smooth and gradual image warping with multi-anchor points’ accurate interpolation.