Abstract: As one of the optimal Non-Destructive Testing（NDT） technology，without any contact，industrial CT system can detect the defect of the object quickly and accurately.In recent years，particularly the emergence of Cone-Beam CT（CBCT），the testing can be better executed according to the three-dimensional isotropic images generated by CBCT.However，the current 2D image enhancement algorithms can only suppress the noise inside the slice，have no access to the noise between slices，can not meet enhancement requirement of the 3D CT images.According to the characteristic of cone-beam CT images，a 3D adaptive smooth algorithm is proposed in this paper.The results show that，with the introduction of the information between slices，the noise can be suppressed further without loss of the edge information，and the performance of this algorithm is obviously better than 2D smooth algorithm.

Key words: Industry Computed Tomography（ICT）, 3D image, image enhancement, adaptive smoothing, discontinuity

摘要： 工业CT系统作为最佳的无损检测技术之一，可以在非接触的条件下快速、准确地发现被检测物体的缺陷部位。尤其近年来新出现的锥束CT，可以生成各向同性的三维图像，更好地对物体进行检测。但是，目前主流的二维图像增强算法，仅能抑制三维图像的层内噪声，无法利用层间信息对图像进一步降噪，尚不能满足三维CT图像增强的需求。根据锥束CT图像特点，将现有的二维自适应平滑算法加以改进，提出了一种三维自适应平滑算法。结果表明：该方法可以利用层间信息，进一步抑制噪声，同时不损失图像的边缘信息，性能明显优于二维平滑算法。

关键词: 工业CT, 三维图像, 图像增强, 自适应平滑, 不连续性