A Perceptual Visual Feature Extraction Method Achieved by Imitating V1 and V4 of the Human Visual System

Abstract

In this paper, we present a new shape encoding method for object recognition. We first introduce a neurophysiologically inspired visual part detector and shape encoder. The optimal form of the visual part detector is a combination of a circular symmetry detector and a corner-like structure detector. A perceptually novel shape descriptor, known as the curvature-orientation descriptor, is then discussed. This descriptor encodes the curvature as well as the dominant orientation. The perceptual shape encoder enhances the performance of feature matching and object recognition taken from standard test images. The results from the repeatability and object recognition tests validate the feasibility of the proposed perceptual feature extraction method.