摘要
在对Canny的最佳边缘检测算子进行理论和实验研究的基础上,针对传统Canny算法在高斯滤波及梯度幅值计算上的缺陷,提出了用中值滤波代替高斯滤波,并通过计算像素八邻域内一阶偏导数有限差分代替2×2邻域内求有限差分来确定梯度幅值的方法,提高了Canny算子在提取图像边缘细节及缓变边缘信息和抑制假边缘噪声方面的性能。实验结果表明,该方法具有较高的信噪比及边缘定位精度,能达到理想的边缘检测效果。
Based on the theoretical and experimental study of Canny＇ optimal edge detection operator,in allusion to the deficiencies of traditional Canny operator on Gauss filtering and grads value account,an improved method,with median filtering instead of Gauss filtering and with the first-order derivative within eight neighborhoods instead of four to count grads value,was proposed.The performance was improved in detecting edge details and slowing changing edges and suppressing noises.Experimental results showed that the improved algorithm had high SNR,high edge orientation accuracy and an excellent edge detection result.
出处
《重庆理工大学学报（自然科学）》
CAS
2010年第5期54-58,63,共6页
Journal of Chongqing University of Technology：Natural Science
基金
重庆市教委科技资助项目(KJ070610)
