一种简化卷积模板的抗噪型边缘检测算法  

An edge detection algorithm of suppression noise based on simplified convolution template

作　　者：沈德海[1] 鄂旭[1] 张龙昌[1] 

机构地区：[1]渤海大学信息科学与技术学院,辽宁锦州121013

出　　处：《信息技术》2015年第6期5-7,11,共4页Information Technology

基　　金：辽宁省高等学校实验室项目(L2012397);博士后基金项目(2012M520158);辽宁省"百千万人才工程"资助项目(2012921058);教育厅科研一般项目(L2012400)

摘　　要：针对传统Sobel算法在边缘定位精度不高、抗噪性能差以及提取边缘较粗等不足,提出一种简化卷积模板的抗噪型边缘检测算法。算法定义了水平方向、垂直方向、45°方向和135°方向的四个简化卷积模板计算图像梯度。在计算方向梯度时,先对参与梯度计算的像素点采用阈值法进行脉冲噪声判断,将灰度值在设定阈值范围内的点视为噪声点,采用3×3窗口进行中值滤波,然后参与梯度计算,对于非噪声点,用其原值计算梯度;对获得的梯度图像进行细化处理并提取边缘图像。仿真实验表明,文中算法提取的图像边缘较细、定位精度较高,而且对脉冲噪声具有较强的抑制能力,图像整体清晰、噪声边缘较少。算法在边缘检测效果及噪声抑制能力上均优于传统的边缘检测算法及小波模变换算法。Aiming at shortcoming of the traditional Sobel algorithm in lower location accuracy, poor noise suppression and coarse edge,this paper proposed an edge detection algorithm of suppression noise based on simplified convolution templata The algorithm defined four simplified convolution templates of horizontal direction,vertical direction,45°direetion and 135°direction to calculate image gradient. While calculating direction gradient, it makes impulse noise judgment for the point participating in calculation by threshold method,takes the points with gray value in the range of threshold as impulse points,and uses median fiher with 3 × 3 window to obtain their new pixel values For non-noise points, it uses their initial value to calculate direction gradient We refine the obtained gradient image, and then extract the image edge The simulation experiments show that the new algorithm has high location accuracy,the edge is fine,and has strong suppressing noise performance; the image is clear,has less noise edge.The new algorithm is better than the traditional edge detection algorithm and wavelet transform algorithm in edge effect and suppression noise ability.

关 键 词：边缘检测 简化卷积模板 阈值 中值滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]