基于改进Canny算法的物体边缘检测算法

Object Edge Detection Algorithm Based on Improved Canny Algorithm

针对传统边缘检测方法受高斯噪声、椒盐噪声污染及边缘梯度变化幅度小等因素影响而出现的物体轮廓检测效果不理想、误检率、漏检率高等问题,提出Canny-Cauchy边缘检测算法。该算法是Canny算法的一种改进,首先对椒盐噪声图像进行自适应中值滤波预处理,在清除椒盐噪声的同时保护边缘不被模糊。在滤波器的设计上,该算法使用柯西分布函数的一阶导数作为边缘检测函数,通过对函数采样得到边缘检测滤波器。对所提边缘检测函数按照边缘检测算法的三条设计准则进行理论分析,并在BSDS500数据集上与其他边缘检测算法进行对比实验。实验结果表明:在降噪方面,该算法可以在20%密度的椒盐噪声下保证处理后图像的峰值信噪比大于30 dB,结构相似性大于0.9;在边缘检测方面,该算法比传统Canny算法对白噪声的抑制能力以及对真实边缘的响应能力更强,在BSDS500数据集上的F1分数提升了7.5%,平均准确率提升了10.2%。

In this study,a Canny-Cauchy edge detection algorithm is proposed to address the issues of unsatisfactory object contour detection performance,high false detection rate,and high missed detection rate of traditional edge detection methods caused by factors,such as Gaussian noise,salt and pepper noise pollution,and small edge gradient changes.The proposed algorithm is an improved Canny algorithm that performs adaptive median filtering preprocessing on salt and pepper noise images to remove salt and pepper noise while protecting edges from blurring.For designing the filter,the algorithm uses the first derivative of the Cauchy distribution function as the edge detection function and obtains the edge detection filter by sampling the function.Theoretical analysis is conducted on the proposed edge detection function according to the three design criteria of edge detection algorithms,and comparative experiments are conducted with other edge detection algorithms on the BSDS500 dataset.The experimental results show that this algorithm can ensure that the peak signal-to-noise ratio of the processed image is greater than 30 dB and the structural similarity is greater than 0.9 under 20%density salt and pepper noise.In addition,this algorithm has a stronger ability to suppress white noise and respond to real edges than the traditional Canny algorithm.Moreover,regarding the BSDS500 dataset,the proposed algorithm exhibites an increase in F1 score and average accuracy by 7.5%and 10.2%,respectively.