一种新的图像边缘检测方法

A New Image Edge Detection Scheme

针对基于空间域或小波变换域图像边缘检测算法提取的边缘只是具有有限方向的问题,该文提出使用具有抛物线型尺度和足够方向消失矩的Contourlet变换来更有效地表示自然图像中的奇异曲线。Contourlet子带比小波子带具有更强的方向性,在其子带上检测系数模极大值,运算复杂度更低。实验结果表明,与基于小波模极大值的图像边缘检测方法相比,该文算法有较低的计算复杂度,所提取的边缘更加逼近图像真实边缘。

To solve the problem of capturing only limited directional information in image edge detection method based on space domain or wavelet transform domain, the Contourlet transform with parabolic scaling and sufficient directional vanishing moments is used to represent discontinuous curves in nature images more efficiently. The algorithm detecting modulus maxima in Contourlet subbands has lower computation complexity, because Contourlet subbands obtain much more directions than wavelet subbands. Experimental results demonstrate that the proposed scheme is superior to the scheme based on wavelet modulus maxima by approximating better the edges of images.