Steerable pyramid filters for selective image enhancement applications

Abstract

In this paper we describe a method we developed for selective detection and enhancement of image feature patterns. A set of directional basis filters were designed to detect the image features of interest, and a steerable pyramid transform is used to generate a multiscale representation of the image features that have desirable shift and rotation invariance. By analyzing and selectively processing the transform coefficients, image feature detection and enhancement can be achieved with designed flexibility in scale, orientation, location, and degree of enhancement. Unlike traditional techniques, this new method promises the selectivity and flexibility desirable in many image enhancement applications. Preliminary experimental results are presented to exemplify the feasibility of the approach.