Detection of anomalies in texture images using multi-resolution random field models

Abstract

In this paper, we present a multi-resolution random field model (RFM) and a corresponding algorithm for anomaly subspace detection. We utilize the redundant discrete wavelet transform (RDWT) for generating a multi-resolution feature space, and model each layer by a non-casual RFM with different sets of parameters. A multi-resolution matched subspace detector (MSD) is designed for detecting targets in the background multi-resolution RFM noise environment. The improved performance of the proposed algorithm is demonstrated compared to using an MSD-based anomaly detector and multi-resolution Gaussian Markov random field (GMRF) model.