一种改进的小波模极大值杂草图像边缘检测算法

An Improved Method of Detecting the Weeds Image Edge by the Modulus Maxima of Wavelet Transform

提出了一种改进的、基于小波多尺度和多分辨率特征的数字图像边缘检测算法,分别利用不同尺度的、小波变换后的、水平方向和垂直方向的高频信息,根据李氏指数与小波变换关系,采用小波模极大值在不同尺度下传播的特性,检测出图像在两个方向的极大值,然后利用模糊算法构造相应的隶属函数,提取弱边缘信息,最后得到不同尺度下的边缘图像。通过该算法对杂草图像仿真实验和经典的两种算子对比,证明该算法可以兼顾良好的边界定位、噪声抑制和弱边界检测等性能指标,能有效地解决传统边缘检测方法中存在的精确及强去噪能力之间的矛盾。

An improved algorithm of numerical image edge detection based on multi-scale and multi-resolution characterization of wavelet transform has been described. This algorithm separately utilizes high frequency information of horizontal and vertical directions of wavelet transform under different scales. According to the relationship of Lipschitz exponent and the wavelet transform ,the properties propagation under different scales are utilized to determine local maxima in two directions. Membership functions are designed with a fuzzy algorithm to detect weak edge information . Finally, all images of different scales are gotten. The algorithm can resolve the inconsistency between high precision localization and high de-noise ability existed in traditional edge detection algorithm.