基于亚像素的边界检测与提取算法及误差分析

The Algorithm and Accuracy Analysis of Edge Detection and Extraction Based on the Sub-Pixel in Digital Image

利用数字图像处理技术已成为解决复杂产品样件反求设计与再设计问题的重要辅助手段,而针对复杂零件 ICT(IndustrialComputedTomography)切片图像的边界轮廓检测和提取精度对反求建模的精度影响至关重要。本文 的主要目的是针对有噪声信号影响的数字图像,在比较现有边界检测与提取算法的基础上,提出了基于亚像素精度 的边界监测与提取算法。为了对该方法的精度进行分析,设计了一个可进行微量驱动控制的工作平台,通过记录工 作平台的移动量和通过边界检测与提取算法得到的位移量之间进行比较的方法分析其精度。通过对误差量使用统 计的方法进行分析,证明了该算法不但能够达到1/4像素的亚像素精度,而且还有计算效率高和计算机开销量小的 特点。

It is one of the important methods to use the technology of industrial computer tomography(ICT) and digital image processing to realize redesign model for complex products in reverse engineering. The accuracy of the location of features in ICT slices affects the quality of the reverse engineering very strongly. In comparison with existing Laplacian and linear interpolation edge-finding method, this paper introduces a new algorithm to detect the location of characteristics and to extract the coordinates based on sub-pixel in an image with blurring and noise. In order to analyze the accuracy of the method, a sample is placed on a drivable test-bed and the differences between the driving-values of the test-bed and the computational displacements by using this algorithm from a lot of CCD digital images are statistically compared. The experiment is carried out to prove this approach to be provided not only with a quart of sub-pixel accuracy for edge detection and extraction, but also with high efficiency and little computational expense.