摘要： 矩方法采用二维理想边缘模型描述亚像素边缘,实际图像在边缘处存在一个渐变的过渡阶段,而二维理想 边缘模型不能精确描述边缘,导致原理误差的产生。为此,提出一种采用误差校正表的算法,用于降低二维理想边 缘模型引入的原理误差,提高亚像素边缘检测精度。通过方形采样定理模拟生成已知边缘的理想图像,用矩方法 检测理想图像的亚像素边缘,构造二维误差校正表。使用查询误差校正表并结合双线性插值求出误差后进行误差 校正。以灰度矩和Zernike 矩为例进行对比实验,结果表明,该算法亚像素检测精度比校正前提高了一个数量级。校正算法主要计算量是双线性插值,能保持亚像素边缘检测的实时性。

关键词: 亚像素, 边缘检测, 矩, 理想边缘模型, 误差校正表, 双线性插值

Abstract: Method of Moments uses two-dimensional ideal edge model to describe sub-pixel edge,but the actual image exists a gradual change on the edge. Thus leads to a principle error that the two-dimensional ideal edge model cannot accurately describe the edge. This paper presents an algorithm which uses an error correction table to regulate error. This algorithm can reduce the principle error from the two-dimensional ideal edge model. And it improves the accuracy of subpixel edge detection. A two-dimensional error correction table is constructed after detecting an ideal image’s sub-pixel edge by method of moments,and simulating the known edge of an ideal image by square sampling theorem. Error correction can be completed after computing the error by querying the error correction table combined with a bilinear interpolation. In a comparison test of gray moments and Zernike moments,experimental results show that compared with methods without regulation this method can effectively improve the accuracy of sub-pixel detection. For the main computational correction algorithm is bilinear interpolation,this method can maintain the real-time sub-pixel edge detection.

Key words: sub-pixel, edge detection, moment, ideal edge model, error correction table, bilinear interpolation

中图分类号: 

• TP301. 6