Tian Tian
China University of Geosciences, China
Title: A biologically inspired spatio-chromatic feature for color object recognition
Biography
Biography: Tian Tian
Abstract
Statement of the Problem: Color information has been acknowledged for its important role in object recognition and scene classification. How to describe the color characteristics and extract combined spatial and chromatic feature is a challenging task in computer vision. To deal with the description of colors, plenty of approaches have been proposed. To study our visual system and mimic its structure and mechanism has naturally been proposed in the research of perceiving colors.
Methodology: The color information processing is implemented under a biologically inspired hierarchical framework, where cone cells, single-opponent and double-opponent cells are simulated respectively to mimic the color perception of primate visual system. More detailed, so responses are obtained by linear combination of retinal cone-like responses generated by Gaussian functions. The receptive field of a do cell is seen as the overlap of two oriented cells with inverse phases, and the double-opponent channels are simulated with oriented filters of inverse phases. Then the robust SIFT feature is extended as a shape description on the processed opponent color channels to obtain a spatio-chromatic descriptor for color object recognition.
Findings: The biologically inspired method is tested for color object recognition task on two public datasets, and the results support the potential of the proposed approach.