Estimating surface characteristics and extracting features from polarisation

Abstract

In this paper we develop a practical method for estimating shape, color and reflectance using only three images taken under polarised light. We develop a novel and practical framework to optimise the estimates and eliminate the redundant information, then investigate three different methods to compare their class discriminating capacities. We present experiment to demonstrate the validity of the proposed method for a database of fruit objects from 5 different classes, and we show that the proposed method is capable of accurately extracting the features of the input examples. The framework can further be applied in a variety fields of computer vision and pattern recognition domains including object recognition and classification. © 2012 Springer-Verlag Berlin Heidelberg.