用于线纹显微图像的边缘检测算法

Edge detection algorithm for lines on microscopic image

为了在简化计算的同时达到较高的定位精度,提出一种轴向邻域和差边缘检测算法,用于低信噪比、缓慢过渡的微结构显微图像的边缘检测。首先,结合显微图像采集系统的配置,分析了线纹显微图像边缘灰度轮廓特征和基于导数的边缘检测算法的不足。然后,基于方向信息测度,定义了轴向邻域和差运算,依据矩不变理论推导出轴向邻域和差边缘检测算法。实验结果表明:轴向邻域和差边缘检测算法能够适应不同分辨率的显微图像,具有较强的抗噪能力和较高的定位精度,边缘检测效果优于基于导数的算法。该算法用于显微图像时,其边缘坐标定位方差为0.57pixel,微米级线条宽度的测量结果与扫描电子显微镜的测量结果(1.35μm)相差0.17μm,基本满足了测量精度的要求。

To simplify computation and to achieve higher positioning accuracy,a Sum and Difference of Neighborhoods along Axis（SDNAA）algorithm was proposed to implement the edge detection of a microscopic image with a lower signal-to-noise ratio and lower transition.Firstly,characteristics of gray profile of the line microscopic image was analyzed and the inadequacy of derivative based edge detection was analyzed according to the configuration of the microscope imaging system.Then,the computation was defined for the SDNAA algorithm based on the orientation information measure,and the SDNAA edge detection algorithm was derived according to the moment invariant theory.Experimental results indicate that the SDNAA algorithm adapts to different resolution images.It has strong anti-noise ability and high positioning accuracy,which is superior to that of the derivative based edge detection algorithm.Using the algorithm for microscopic images with low signal-to-noise ratio and low contrast,the variance of located edge coordinates is 0.57 pixel,and the measurement result of micro line width deviates from that of SEM （1.35 μm） by 0.17 μm,which reaches the expected precision.