A NEW APPROACH FOR SUBSET 2-D AR MODEL IDENTIFICATION FOR DESCRIBING TEXTURES

This paper addresses the problem of identification of appropriate auto regressive (AR) components to describe textural regions of digital ima ges by a general class of two-dimensional (2-D) AR models. In analogy with univariate time series, the proposed technique first selects a ne ighborhood set of 2-D lag variables corresponding to the significant m ultiple partial auto-correlation coefficients. A matrix is then suitab ly formed from these 2-D lag variables. Using singular value decomposi ton (SVD) and orthonormal with column pivoting factorization (QRcp) te chniques, the prime information of this matrix corresponding to differ ent pseudoranks is obtained. Schwarz's information criterion (SIC) is then used to obtain the optimum set of 2-D lag variables, which are th e appropriate autoregressive components of the model for a given textu ral image. A four-class texture classfication scheme is illustrated wi th such models and a comparison of the technique with a recent work in the literature has been provided.