摘要： 由于能提供较多的方向信息，双树复小波变换在纹理图像检索中的检索率高于传统小波变换，但传统小波变换与双树复小波变换得到的方向子带不同。针对该问题，提出一种融合传统小波和双树复小波变换的一阶统计信息从而提取特征进行纹理图像检索的方法。对Brodatz图像库的仿真实验表明，该方法优于传统小波和双树复小波方法。

关键词: 纹理图像检索, 复小波变换, 传统小波变换, Canberra距离

Abstract: Texture retrieval with Dual-Tree Complex Wavelet Transform(DTCWT) is effective to improve performance relative to conventional wavelet transform due to its multi-directional information. The conventional wavelet transform gives different directional information from DTCWT. Aiming at this problem, this paper proposes a texture image retrieval approach combining the DTCWT and the conventional wavelet transform to extract features. Experimental results on Brodatz database show that the proposed method has better performance than the DTCWT and the traditional wavelet transform.

Key words: texture image retrieval, complex wavelet transform, conventional wavelet transform, Canberra distance

中图分类号: 

• TP391