Nonlinear Image Interpolation Through Extended Permutation Filters

This paper considers the task of image enlargement, or interpolation. It takes a nonlinear filtering approach to interpolating missing values in the enlarged target image. It is based on a generalization of the median filter known as a rank selection filter. These filters are constrained to output an order statistic based on input values within a sliding observation window. In particular, we allow the filter to output arbitrary linear statistics extracted from the sliding window. We show that these filters produce superior results to traditional methods for image enlargement.