基于改进Canny算子的含噪图像边缘检测  

Edge detection method for noisy images based on improved Canny operator

作　　者：李萍[1] 杨丹[2] LI Ping;YANG Dan(School of Information Engineering,Tongren Polytechnic College,Tongren 554300,China;School of Computer and Information Science,Chongqing Normal University,Chongqing 401331,China)

机构地区：[1]铜仁职业技术学院信息工程学院,贵州铜仁554300 [2]重庆师范大学计算机与信息科学学院,重庆401331

出　　处：《邵阳学院学报(自然科学版)》2024年第6期20-27,共8页Journal of Shaoyang University:Natural Science Edition

基　　金：重庆市社科规划项目(2022NDYB119)。

摘　　要：针对含噪图像的边缘检测较低、无法准确提取出真实边缘,导致图像视觉效果较差的问题,提出基于改进Canny算子的含噪图像边缘检测。改进Retine对图像进行增强,获取增强后的反射图像;经各向异性扩散滤波法对增强后图像进行去噪和平滑处理;优化Canny算子,采用Canny算子中梯度幅值双阈值作为判定依据,对去噪后的平滑图像进行边缘检测,提升图像边缘检测效果。实验结果表明,该方法对含噪图像去噪后,信噪比为57.96,均方根误差为3.12e-04,噪声指标从1左右降到了0.13左右,增强后图像有效解决了光晕现象并可实现暗区域增强,边缘检测不存在间断,且无虚假边缘。由此说明,应用该方法实现良好图像去噪的同时,保证了图像边缘锐度,提高了图像视觉效果。An edge detection method was designed explicitly for noisy images utilizing improved Canny operators to tackle the issue of inadequate visual effects in images resulting from low-quality edge detection and failure to accurately extract the real edge in noisy conditions.This study used an improved Retine algorithm to extract enhanced reflection images,which were then denoised and smoothed using an anisotropic diffusion filtering technique.Edge detection was carried out on the smoothed images based on gradient amplitude dual threshold from the optimized Canny operators,significantly improving edge detection performance.Experimental results demonstrated that after applying this denoising method,the images achieved a signal-to-noise ratio of 57.96,a root mean square error of 3.12e-04,and a reduction in the noise index from approximately 1 to around 0.13.The enhanced images displayed no halo effects but improved clarity in darker regions,with no interruptions in edge detection or false edges.These findings suggested that this approach effectively achieves good image denoising while maintaining sharpness in edges,thereby enhancing the overall visual quality of the images.

关 键 词：图像去噪 各向异性扩散滤波 改进Canny算子 边缘检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]