In order to detect circles automatically and to obtain the parameters in images precisely, a method to measure circles in industrial CT images was developed. Firstly, the sub-pixel edge of an image was detected based on a Facet model. Then, a method for calculating the existence probability map of circles and its defects were analyzed. To overcome the defects of long-time calculation and memory consumption, the present method made sure the potential range of the center of a circle and the stored message, which met the required numerical values in a chain. Finally, the circles were detected based on the improved method, and the circle parameters were calculated with Least Square Algorithm(LSA). The method was applied to measure an image with 10 circles obtained from an industrial CT system with the spatial resolution of 2.0 lp/mm. The results show that this method has greatly improved the calculation rate, and the measurement accuracy for circles is better than 0.5%. It can satisfy the measurement requirements of the circles for the higher speed, higher precision and non-contact.