Content-based image retrieval using combined texture and color features based on multi-resolution multi-direction filtering and color autocorrelogram

This paper proposes content-based image retrieval using combined feature extraction method by local energy, rotation-invariant uniform local binary patterns (RULBP) and color autocorrelogram. In this paper, features of local energy and RULBP are extracted in a high frequency domain of a gray image, using multi-resolution multi-direction (MRMD) filtering. Autocorrelogram is used in a two-dimensional histogram of hue and saturation color spaces. The Mahalanobis distance is used in measuring the degree of similarity of features. Evaluation of the experimental result of the image retrieval is based on precision and recall. The tested DBs (databases) are Corel and VisTex; Corel_MR and VisTex_MR of resolution variants; and Corel_MD and VisTex_MD of direction variants of first two DBs, respectively. The result of the experiment shows that the proposed method is superior to compared methods in the aspect of retrieval performance to retrieve MRMD images.