Soble-Zernike矩亚像素边缘算子的影响因素分析

针对影响Sobel-Zernike矩算子运行速度、定位精度的有关因素进行了讨论。采用平均一自适应联合的平滑算法，显著改善了边缘定位精度，其实质在于消噪和锐化边缘。推导出边缘点新、旧判据amp和k间的关系式，在，取值较小时，k／amp恒等于1．5，可见amp对检测精度的影响可以忽略。分析结果表明，当图像中边缘点比率为87．75％时，Sobel-Zernike矩和Zernike矩算子的运行时间相等，由于实际图像中边缘点所占比率不会超过50％，故前者的运行时间总是小于后者。

Analysis of the influencing factors on subpixel Sobel-Zernike moments edge operator

Sobel-Zernike moments operator has some advantages of both runtime and location accuracy. For well understanding this algorithm, some influencing factors on location accuracy and runtime are discussed in the paper. Test results show that average-adaptive smoothing operator can evidently improve location accuracy. The reason is that the combined smoothing operator can eliminate noise and make edge sharp. The relationship between new criterion amp and old criterion k is obtained by theoretical analysis. The ratio k/amp is equal to 1.5 when edge parameter l is predetermined as a small value and the influence of amp on location accuracy to judge edge can be ignored. The calculating result shows that the runtime of Sobel-Zernike and Zernike moments operator is equal at 87.75% of edge points. Since the ratio of edge points is low than 50% in real image, the runtime of Sobel-Zernike moments operator is always shorter than that of Zernike moments operator.