一种利用非下采样Contourlet变换的纹理检索方法

Novel approach for texture retrieval using nonsubsampled Contourlet transform

为提高基于内容的图像检索系统中纹理特征提取的有效性,提出了又一种纹理图像检索方法。该方法利用非下采样Contourlet变换对图像进行分解,提取不同子带和不同方向变换系数矩阵的均值和方差为特征向量,作为数据库中纹理图像的索引,并利用两种不同的相似度函数计算图像之间的相似度,建立了一套基于示例查询图像的纹理图像检索系统。实验结果表明,与小波包等特征提取方法相比,该方法不仅能降低特征向量维数,而且能取得更高的检索准确率和检索速度。

To increase the validity of texture feature extraction in content-based image retrieval system, a novel approach for texture image retrieval was proposed. This approach was based on the NonSubsampled Contourlet Transform （ NSCT）. The means and variables of NSCT coefficients matrix in different subbands and various directions were extracted to form the feature vectors which were regarded as indexes of texture images in image database. Two similarity functions were used to compute the similarity between images. A texture retrieval system based on query image was developed. Compared to the wavelet package transform, this approach can not only reduce the dimension of feature vectors, but also get higher accuracy and speed of retrieval.