Texture Image Retrieval Based on NSCT and DT-CWT

Abstract

Abstract This paper presents a novel algorithm for texture image retrieval based on nonsubsampled contourlet(NSCT) and Dual-tree complex wavelet transform(DT-CWT). Nonsubsampled contourlet transform has anisotropy and translation invariability. DT-CWT not only has good localization in time-frequency domain, but also has approximate translation invariant, more directivity and limited data redundancy. Texture features based on NSCT and DT-CWT can include more information of image. This paper also presents a novel similarity measure. Experimental results demonstrate that the proposed approach improves average retrieval accuracy.