Texture Image Retrieval Based on Rotated Complex Wavelet Signatures

Abstract

In this paper, we propose a new approach of extracting texture features for texture image retrieval. A set of rotated wavelet filters (RCWF) is used jointly with dual-tree complex wavelet transform (DT-CWT) for texture analysis in twelve different directions. Wavelet signatures are generated from the decomposition coefficients. Graph-based spectral clustering method is applied on the distance of the signatures. Because the wavelet signature represents the image texture well and the spectral based method exploits the local relationship significantly, experimental results show that the proposed method outperforms other existing methods remarkably.