PHLST with adaptive tiling and its application toantarctic remote sensing image approximation
Zhihua Zhang, Naoki Saito · Inverse Problems and Imaging · 2014
We propose an efficient nonlinear approximation scheme using thePolyharmonic Local Sine Transform (PHLST) of Saito and Remy combinedwith an algorithm to tile a given image automatically and adaptivelyaccording to its local smoothness and singularities. To measure suchlocal smoothness, we introduce the so-called local Besov indices ofan image, which is based on the pointwise modulus of smoothness ofthe image. Such an adaptive tiling of an image is important forimage approximation using PHLST because PHLST stores the corner andboundary information of each tile and consequently it is wasteful todivide a smooth region of a given image into a set of smaller tiles.We demonstrate the superiority of the proposed algorithm usingAntarctic remote sensing images over the PHLST using the uniformtiling. Analysis of such images including their efficientapproximation and compression has gained its importance due to theglobal climate change.