【摘要】 线性特征是图像的一种重要局部特征,它常常决定图像中目标的形状。线性特征的提取在图像匹配、目标描述与识别以及运动估计、目标跟踪等领域具有十分重要的意义。常用的线性特征检测方法有Radon变换和Hough变换,但检测曲线复杂度会很高。本文提出一种多尺度几何分析的线性特征检测方法,该方法以finite ridgelet理论为基础,结合正交小波变换对线性特征进行提取。Finite ridgelet变换对于含有直线奇异的多变量函数具有良好的逼近特性,能够获得连续空间函数的稀疏表达,同时具有区域平滑性、很好的可逆性和去冗余性。实验结果表明,本方法即使在背景复杂的环境下也具有良好的检测效果。更多还原

【Abstract】 Linear feature detection is very important in image processing and frequently uses in object recognition,image alignment,image matching,movement estimation and object tracking and so on.The Radon and Hough transform is most commonly used for the detection of regular curves such as lines,circles,ellipses,etc.But they are only fit for images with regular shape and only when linear feature is distinct,the lines in the image are efficiently detected.On the other hand,the length of a line cannot be kept by these two transforms,only the position of the line can be determined.And the computational complexity is very high especially in detecting circle features.Finite ridgelet transform is a discrete orthonormal version of ridgelet transform and based on finite ridgelet transform,a more robust linear feature extraction approach combined with dyadic wavelet transform is proposed.Experimental results show the efficiency of this algorithm even in noised scene.更多还原