Registration of Airborne Scanner Imagery Using Akima Local Quintic Polynomial Interpolation

Evaluation and analysis of most remotely sensed data requires careful image-to-image registration or ortho-rectification (geocoding). Photogrammetric practice has shown that polynomial transformation functions are useful for registration of aerial photographs. However, multiand hyperspectral remotely sensed data are often recorded by airborne line scanners. Then ortho-rectification by conventional global coordinate transforms is not satisfactory, particularly due to the non-instantaneous image formation process. It is rather necessary to allow for local corrections of distortions. We have implemented two interpolation techniques for the computation of the resampling coordinates, local AKIMA and HARDY multiquadric interpolation, and applied them to experimental image data. Comparison shows that resampling based on AKIMA interpolation is much faster than on HARDY multiquadric. Moreover, AKIMA interpolation in the limit corresponds to conventional second degree polynomial transformation. It is thus an appropriate instrument for local image registration utilizing ground control points.