Multispectral texture analysis using interplane sum- and difference-histograms

Abstract

In this paper we present a new approach for color texture classification which extends the gray level sum- and difference histogram features [8]. Intra- and inter-plane second order features capture the spatial correlations between color bands. A powerful set of features is obtained by non-linear color space conversion to HSV and thresholding operation to eliminate the influence of sensor noise on color information. We present an evaluation of classification performance using four different image sets. © Springer-Verlag Berlin Heidelberg 2002.