Dominant and minor sum and difference histograms for texture description

Abstract

In this paper, we introduce a new color texture operator for natural texture classification, the Dominant and Minor Sum and Difference Histograms (DM-SDH) descriptor. The proposed approach allows to incorporate both color and texture information in order to enhance the texture discrimination performance. For this purpose, a vectorial representation of the image is used for the descriptor extraction instead of the scalar representation in addition to the use of the dominant and minor sum and difference to encode the local structure. The performance of the proposed descriptor is evaluated using the New-BarkTex and the Outex-TC13 datasets. The experimental results show that the DM-SDH operator significantly outperforms the state-of-the-art methods.