摘要
应用经典限幅滤波算法(CFA)对边缘去噪处理时容易造成有效的高频边缘被抑制、破坏边缘连续性、丢失目标结构特征的问题。提出了一种基于局部结构形态特征的改进型限幅滤波算法,用于图像边缘滤波处理。该算法利用限幅滤波的原理,引入由5个相邻边缘点构成的滑动模子并遍历边缘各点,对滑动模子中增量超限的点加以结构形态预测和阈值判断,即通过滑动模子建立局部轮廓的结构形态模型,并应用模型进行边缘预测;对超限点与预测值的差异进行了比较,为判定是否遇到台阶、凸缘或尖锐的结构特征提供了依据。为了测试新算法在边缘保持和滤波降噪方面的能力,与传统限幅滤波算法进行了对比实验。实验结果表明:基于局部结构形态特征的改进限幅滤波算法不但具有高效的去噪能力,而且对目标结构中的高频边缘具有显著保护作用。
It is easy to cause the problem that the high frequency edges are suppressed,the continuity of the edges is destroyed and the structural features of the target are lost when the classical clipping and filtering algorithm(CFA)is applied to the edge denoising.Therefore,an improved CFA based on local structure shape features is proposed for image edge filtering.Based on the principle of CFA,a sliding template composed of 5 adjacent edge points and sliding along the edge is used in the algorithm.The point that overruns the threshold in the sliding template is predicted by the local structural shape model.To be specific,the model of the local structural contour is established by using the sliding template and applied to predict the overrun point.Then,it is judged that whether the difference between the overrun point and its predicted value is greater than the threshold,in the physical sense it means that whether the step,flange or sharp structural features are encountered.In order to test the ability of the new algorithm in edge preservation and noise reduction,comparative experiments with the traditional algorithmare carried out.The experimental results show that the improved CFA based on local structure features not only has efficient denoising ability,but also has significant protection effect on the high-frequency edge of the target structure.
作者
孙晓辉
蔡永洪
林雁飞
SUN Xiao-hui;CAI Yong-hong;LIN Yan-fei(Guangzhou Institute of Technology,Guangzhou,Guangdong 510075,China;Guangzhou Institute of Measure and Testing Technology,Guangzhou,Guangdong 510663,China)
出处
《计量学报》
CSCD
北大核心
2022年第1期21-25,共5页
Acta Metrologica Sinica
基金
广东省市场监督管理局科技计划项目(2020CJ02)
广东省教育厅2020年省高职教育教学改革研究与实践项目高职扩招专项(JGGZKZ2020114)。
关键词
计量学
图像边缘滤波
限幅滤波算法
局部结构形态
机器视觉
metrology
image edge filtering
clipping and filtering algorithm
local structure shape
machine vision
