摘要
为了满足计算机视觉标定与精密测量对图像边缘定位的高精度和抗噪声要求,提出一种基于正交Fourier-Mellin矩亚像素边缘检测的改进算法.首先,建立亚像素边缘理想模型,推导了新的Fourier-Mellin矩7×7模板,利用各级矩的卷积来提取边缘点细节特征.然后,根据幅值旋转不变性原理,通过分析图像边缘旋转至垂直方向后各阶次Fourier-Mellin矩之间的关系,得到被测点亚像素边缘的各级参数.再根据改进的边缘判据,利用其高阶矩的细节描述能力和低阶矩的噪声抑制能力,确定图像中实际边缘点的亚像素坐标.实验结果表明,改进算法具有更高的检测精度和更强的抗噪性能,对含噪直线和含噪曲线的定位精度可达0.07像素和0.10像素,更好地满足了图像边缘定位的稳定可靠、高精度测量要求.
An improved suhpixel edge detection algorithm based on orthogonal Fourier-Mellin moments was proposed to meet the stringent requirements for high location accuracies and strong anti-interfer- ence ability in computer vision calibration and precision measurement. The ideal step model of subpixel edge was established, and a new mask of 7×7 Fourier-Mellin moments was calculated. Thus the mi- nutiae of edge points were extracted through the convolution with each order of Fourier-Mellin mo- ments. The relationships between different orders and repetitions of Fourier-Mellin moments were an- alyzed when the image edge was rotated to the vertical direction. According to the principle of amplitude rotation invariance, the parameters of subpixel edge were obtained. And then, a more reasonable criterion for edge detection was put forward by making full use of the merits of higher order moments in detail extraction and the lower order moments in noise inversion, to make the edge detection results better. The experimental results indicate that the proposed method has the virtue of higher measuring precision and better noise suppression performance. The edge detection accuracy is up to 0.07 pixel for straight lines with noise, and 0.10 pixel for curves with noise. Therefore, the improved algorithm is an efficient approach with a relatively high accuracy and stabilization for image edge detection.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2010年第5期53-56,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
中国博士后科学基金资助项目(20080430965)
作者简介
张坤（1983-），男，博士研究生，E-mail：jatity@163．com．
