摘要
Nonsubsampled Contourlet变换是一种非抽取得具有平移不变性的多尺度多方向的变换。将Canny算子和Nonsubsampled Contourlet变换(NSCT)[1]相结合,对图像运用Canny算子[2]提取边缘特征,再进行Nonsubsampled Contourlet变换,引入了三阶中心矩作为特征向量提取形状特征的算法。实现了基于Nonsubsampled Contourlet变换的图像形状特征检索,并将结果与基于2-D小波变换和基于Contourlet变换的图像形状特征检索作了比较,实验结果证明该方法的图像形状特征检索效率有较大的提高。
The nonsubsampled Contourlet transform （NSCT） provides a flexible multi-scale, multi-direction, and shifi-invariant image representation. A new approach to combine the advantages of both Canny algorithm and NSCT for SAR image retrieval of shape features is proposed in this paper. Firstly the Canny operator to detect the edges of image is used, and then the image is decomposed with NSCT. The energy, standard deviation, and third central moment are choosen as the shape features. The shape retrieval performance of NSCT is compared to those of the conventional transforms including the 2-D wavelet, and the Contourlet. An experiment shows that the NSCT outperforms the other two trans-forms in texture image retrieval.
出处
《雷达与对抗》
2008年第1期25-29,共5页
Radar & ECM
基金
江苏省自然科学基金(BK2001047)
航空科学基金(04D52032)
