为提高工业计算机断层扫描(CT)图像亚像素边缘检测的精度和速度，研究了一种改进的Zernike矩边缘检测方法。该方法采用Sobel边缘算子快速检测出图像所有可能的边缘，通过Zernike矩算子对所有可能的边缘进行重新检测，最后，检测出图像的亚像素边缘并计算其精确位置。由于采用Sobel算子检测出可能的边缘使后续Zernike矩算子检测范围缩小，从而减小了运算量，提高了运算速度。对实际CT图像进行的实验结果表明: 改进的Zernike矩工业CT图像边缘检测精度绝对误差＜0.24 pixel，改进算法的运算速度提高了约70%。

In order to improve the detecting speed and accuracy of sub-pixel edges for an Industrial Computed Tomography(ICT) image, a kind of improved edge detecting method based on Zernike moments is studied. First, the Sobel edge operator is used to quickly detect all the possible edges of the image, and then the Zernike moment operator is taken to re-test all the possible edges. Finally, the exact positions of the sub-pixel edges can be detected and calculated. As the possible edges have been detected and the detection range is reduced, the computing amount is shortened and the operation speed is imporved. An experiment is performed on CT images, and the result shows that the absolute error for the accuracy is less than 0.24 pixel,while Zernike moment operator computing speed is increased by about 70%, which ensures the accuracy of edge detection and improves the speed of edge detection.