尺度及旋转不变纹理图像检索

Image retrieval with scale and rotation invariant texture features

针对图像检索时常见的尺度及旋转问题,本文提出了一种基于双树复小波变换的尺度及旋转不变纹理图像检索方法。对图像进行尺度变化的预处理,对原图像和变化后的图像分别进行DT-CWT后,提取纹理特征并插值为特征矩阵;分别在尺度维和旋转维上进行FFT消除尺度和旋转变换影响,得到尺度和旋转不变的特征向量;采用Canberra距离进行相似性度量。通过对尺度及旋转变化的纹理图像库的实验表明,该方法对图像的旋转和尺度变化具有较好的鲁棒性。

A method used for image retrieval with scale and rotation-invariant texture feature was proposed against the common scale and rotation problems in image retrieval. Texture image was decomposed with the dual-tree complex wavelet transform （DT-CWT）. Firstly, scale the input image with a constant, filter the two images with DT-CWT and extract texture features,and a feature matrix was formed by interleaving the two results of filtering by scaling level of the coefficients. Secondly, compute the FFT of the feature matrix over scale and rotation dimension to eliminate the influence of scale and rotation trans-gorm, and a feature vector which was invariant to scale and rotation is obtained. Then, the similarity was measured by Canberra distance. At last, the experiments carried on scale and rotation texture database can prove that the proposed approach is insensitive to the scale and rotation transform of image.