Abstract

Most widely used geometric correction methods rely on ground control points to empirically determine a mathematical coordinate transformation to project the distorted image into the reference image. However, conventional methods usually employed the same correction formula in the whole image which few considered the seriously geometric distortion problem causing by imaging with a large view angle. This paper introduces a new piecewise polynomial correction method that addresses this problem based on the nonlinear variability of geometric distortions in different imaging angles. In the new method, piecewise strategy and the selection of control points are two key techniques. The seriously oblique image is partitioned into several subspaces which separately use the different correction factors in order to correct the local distortions better. And in the correction, a uniform distribution strategy is proposed for selecting the representative control points to evaluate the correction polynomial coefficients of each subspace. The experimental results show that the proposed correction method is significantly outperforming conventional approaches. For an image acquired in 60 degrees, the proposed method results in correction root mean square errors having around 30% improvement comparing to the traditional method.

Paper Details