摘要
除具有纹理属性外,遥感图像中还包含大量的结构性边缘特征。如何有效捕捉这些特征中的信息进行检索,成为提高遥感图像检索效率的关键。依据Contourlet在离散域的多尺度几何分析的功能,提出一种利用Contourlet子带能量分布特性提取原始图像在多个尺度下的边缘方向信息进行检索的方法。针对Cont-ourlet变换响应不同特征方向能力的差别,采用正交补偿法加以改正,经过傅里叶算子的处理,从而实现旋转不变结构性边缘特征的检索。最后,在遥感目标图像上所作的实验证实了该算法的有效性。
Besides texture features, remote sensing images process abundant structured edge features. The main problem here is how to efficiently capture the information contained in these features to facilitate image retrieval effectively. Based on the multi-scale geometric analysis function of Contourlet, a kind of retrieval technique has been addressed, which uses the information of edge direction extracted from the multiple scale sub-bands of Contourlet through decomposing the original image. Against the limitation that Contourlet transformation responds differently to diverse directional features, an orthogonal compensation method has been put forward to revise the flaw. Subsequently, Fourier operator has been applied to realize rotation-invariant image retrieval technique. Finally, retrieval experiments have been performed on remote sensing target images to confirm the effectiveness of the proposed algorithm.
出处
《测绘科学技术学报》
北大核心
2009年第2期148-152,共5页
Journal of Geomatics Science and Technology
关键词
轮廓波
正交补偿
边缘特征
遥感
图像检索
Contourlet
orthogonal compensation
edge features
remote sensing
image retrieval
