Image Retrieval Based on Multi-wavelet and Non-subsampled Contourlet Transform
Zhenqi Niu · Journal of Henan Normal University · 2011
To improve the quality of texture image retrieval,this paper proposes a new method for texture image retrieval based on multi-wavelet and non-subsampled contourlet transform.Firstly,the image is decomposed into multi-scale with multi-wavelet to ensure translation invariant and eliminate transform redundancy.Then,multi-directional decomposition of the high-frequency sub-bands is implemented with non-subsampled directional filter bands in every scale,the first-order geometric distances of the low frequency sub-band and every scale's directional sub-band are calculated as the texture features.Finally,the retrieval results are obtained using the Euclidean distance.The experimental results show that the proposed method's retrieval performance is better than that of the traditional multi-resolution approaches.