摘要
图像的边缘是图像最重要的特征之一。由于边缘和噪声都是图像的高频分量,提取的图像边缘总是受到噪声的污染。针对边缘检测中存在的噪声问题,本文根据Mallat快速小波变换算法的思想,提出用高斯函数和其一阶导数分别作为低通和高通滤波器对图像进行多尺度分析。为了精确定位图像边缘,对各尺度的低频、水平、垂直和对角分量不进行下采样。然后提取不同尺度上的系数,利用多尺度积对噪声严重的图像进行边缘检测。最后根据边缘点的梯度方向,采用改进的局部梯度极大值搜索方法获得图像的单像素边缘。实验结果表明本文所提出的方法,能在被噪声污染严重的图像中提取图像的单像素边缘,且边缘图像信噪比高。
Edge is one of the most important features of image. The extracted edge is always polluted by noise because edge and noise are with high frequency. In order to solve conflicts in suppressing noise and detecting edge, a new edge detection algorithm based on non-sub-sampled and multi-scale product is presented according to Mallat fast wavelet transform algorithm. It uses Gaussian function and its first-derivative as low-pass and high-pass filters to decompose image, respectively. For the location of accurate edge, low-frequency, horizontal, vertical and diagonal information of each scale are not sub-sampled. Then, multi-scale product is adopted to detect edge and suppress noise. Finally, the updating search method of local modulus maximum is used to achieve single-pixel edge by gradient direction of edge pixels. Experiment results show that the proposed algorithm has advantages of detecting single-pixel edge in the image embedded by noise with high signal-to-noise ratio.
出处
《数据采集与处理》
CSCD
北大核心
2012年第4期490-494,共5页
Journal of Data Acquisition and Processing
基金
上海市教育委员会(10YZ171)资助项目
关键词
边缘检测
多尺度积
高斯小波
梯度值
edge detection
multi-scale product
Gaussian wavelet
gradient magnitude
