Abstract: To improve accuracy and efficiency of contour processing of industrial computed tomography (CT) images for reverse engineering, an industrial CT slicing image contour extraction and optimization system is developed in this paper. The system includes five parts, which are modules of image processing, contour recognition and extraction, contour fitting and optimization, image processing algorithm library, and contour data format transform. The processing results of the system can be organized as 3-D point clouds or contour line used standard format, which can be used to reconstruct 3D CAD model directly in professional softwares of reverse engineering. Further more, a fitting model based on global constraints is constructed to find the optimal contour curves, and a method combining directed graph decomposition and numerical algebra is proposed to improve the accuracy and efficiency of fitting contours. Finally, the effectiveness and feasibility of this system is proved with an experiment.