Abstract - This paper presents a new technique to index and retrieve texture image

This paper presents a new technique to index and retrieve texture image. The image is modeled with the two-dimensional one-sided linear prediction (OSP) model. The prediction coefficients, normalized mean-square prediction error and a set of seven invariant moments of the prediction error are used as features to index the image. The proposed method is highly successful in retrieving perceptually similar textures. Index Terms – Content-based image retrieval (CBIR), texture modelling, one-sided linear prediction (OSP), linear-prediction error.