Statistical Approaches to Material Classification

Proceedings of the Indian Conference on Computer Vision, Graphics and Image Processing, Ahmedabad, India |

The objective of this paper is classification of materials from a single image obtained under unknown viewpoint and illumination conditions. Texture classification under such general conditions is an extremely challenging task. Our methods are based on the statistical distribution of rotationally invariant filter responses in a low dimensional space. There are two points of novelty: first, two representations of filter outputs, textons and binned histograms, are shown to be equivalent; second, two classification methodologies, nearest neighbour matching and Bayesian classification, are compared.

In essence, given the equivalence of texton and bin representations, the paper carries out an exact comparison between the texton based distribution comparison classifiers of Leung and Malik [IJCV 2001], Cula and Dana [CVPR 2001], and Varma and Zisserman [ECCV 2002], and the Bayesian classification scheme of Konishi and Yuille [CVPR 2000].

The comparisons are assessed by classifying images of all 61 materials present in the Columbia-Utrecht database. Classification rates of over 97% are achieved for both the methods while classifying more than 2800 images in all.