A metric algorithm based on three elements of texture visual feature

Abstract

In order to construct a reference model to recognize image texture, a combining method based on three elements of texture visual features is proposed. Firstly, a fractal model is used to calculate the fractal dimension which is a measure of image textural coarseness. Secondly, a global texture direction is proposed. Gabor filter and local marginal probability histogram is used to calculate a quantitative value of texture direction. Thirdly, the texture contrast base on Tamura model is applied to describe image texture feature. Finally, the combined method based on the coarseness, the direction and the contrast is applied to extract texture visual features in Brodatz texture database. The experimental result is consistent with human visual perception. The algorithm can be better reference model to satisfy machine identification image texture.