基于旋转不变小波阈值的多尺度边缘检测

在自适应多尺度边缘检测算法中,用旋转不变小波非线性阈值代替原来的高斯滤波做平滑,并对尺度指数采样产生尺度集合,以确定边缘像元的最佳尺度,得到基于旋转不变小波阈值的多尺度边缘检测算法.该算法避免了对原图像的过度光滑,用于检测边缘的最佳尺度与边缘存在的尺度空间范围相匹配.实验证明新算法检测的图像比较清晰,具有较高的信噪比,而且可以用来检测含少量白噪声图像.

Multi-scale Edge Detection Based on Rotationally Invariant Wavelet Shrinkage

A new multi-scale edge detection is proposed in which we replace the Gaussian filter in adaptive multi-scale edge detection with rotationally invariant wavelet shrinkage,and the filter scale set is determined by sampling exponentially,not by equidistantly.The new algorithm avoids the over-smoothness of the original image resulting from the Gaussian filter and the optimal scales of the edge pixels are matching with the ranges of the scale space the original image belongs to.Experiments show the new algorithm obtains a better edge detection result with higher signal to noise ratio, even if the image is contaminated with little white noise.