Classification between natural and graphics images based on generalized Gaussian distributions

摘要

We propose a novel method for classifying photographic and computer-generated images based on generalized Gaussian distribution (GGD) modeling of subband coefficients. The estimated shape and standard deviation parameters of GGD within each resolution level, the ratio of the estimated shape parameters between different resolution levels, and the ratio of the estimated standard deviation parameters between different resolution levels are used as features for the classification. (C) 2018 Elsevier B.V. All rights reserved.