改进的Roberts图像边缘检测算法  

Image Edge Detection Algorithm of Roberts Operator

作　　者：王方超[1,2] 张旻[1,2] 宫丽美 

机构地区：[1]解放军电子工程学院,安徽合肥230037 [2]安徽省电子制约技术重点实验室,安徽合肥230037

出　　处：《探测与控制学报》2016年第2期88-92,共5页Journal of Detection & Control

基　　金：国家自然科学基金项目资助(60972161);安徽省自然科学基金项目资助(1408085QF115)

摘　　要：针对复杂背景下图像边缘检测中存在抗噪性能不强、边缘不连续等问题,提出了改进的Roberts边缘检测算法。该算法采用3×3邻域代替Roberts算法中2×2邻域来计算梯度幅值;并利用图像块之间相似性的三维块匹配的去噪模型,提高Roberts算子的检测精度和抗噪性能;通过最佳阈值迭代方法代替人为指定阈值来获取最佳分割阈值,有效地提取图中目标轮廓。仿真实验结果表明,该算法PSNR达到33dB左右,比抗噪形态学边缘检测算法和一种改进的Roberts和灰色关联分析的边缘检测算法抗噪性能好,在抑制噪声干扰的同时,能保留边缘信息,较好提取目标的整体轮廓信息,为后续目标识别奠定基础。Aiming at the problems of the resist noise performance is poor,the edge is not continuous when detecting the edge of image in the complex background,an improved Roberts edge-detection method was proposed.The algorithm adopted neighbor-field to replace neighbor-field to calculate grades extent value.And based on image block similarity between 3Dblock matching denoising model,detection accuracy and anti-noise performance of Roberts operator was improved.The artificial specified threshold method replaced the optimal iterative threshold segmentation method to obtain optimal segmentation threshold for effectively extract the target contour in image.Simulation results showed that the algorithm with PSNR was superior to the algorithm of morphology in edge detection to resist the noise with an improved image edge detection algorithm based on Roberts and gray relational analysis,which could suppress noise interference and keep the edge information.

关 键 词：边缘检测 梯度幅值 三维块匹配 最佳阈值迭代分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]