Description

We follow the transform-based texture classification paradigm and propose a novel Bayesian texture classifier via the BYY harmony learning of Poisson mixtures on the contourlet features of texture images. Here, we model the contourlet coefficients in each sub band using the simpler probabilistic models, i.e., Poisson mixtures, rather than the relatively complicated HMT model. To establish a novel kind of Bayesian texture classifier through the BYY harmony learning of Poisson mixtures on the contourlet features of texture images. To make use of the contourlet decomposition to set up an efficient representation of the image for texture classification, whereas the BYY harmony learning of Poisson mixtures can lead to a good parameter estimation with adaptive model selection.