Enhancing signal discontinuities with shearlets: An application to corner detection

Shearlets are a relatively new and very effective multiresolution framework for signal analysis able to capture efficiently the anisotropic information in multivariate problem classes. For this reason, Shearlets appear to be a valid choice for multi-resolution image processing and feature detection. In this paper we provide a brief review of the theory, referring in particular to the problem of enhancing signal discontinuities. We then discuss the specific application to corner detection, and provide a novel algorithm based on the concept of a cornerness measure. The appropriateness of the algorithm in detecting good matchable corners is evaluated on benchmark data including different image transformations.