Abstract: It is a key technique for computer vision and image understanding to detect and measure circular objects in images. Existing algorithms possess several drawbacks such as low detecting and measuring accuracies and poor robustness to image degradation. This paper proposes a two-stage hybrid approach to solve this problem. In the first stage, a modified Hough transform is used to obtain the accumulative space of the original image; in the second stage, connected components in the horizontal cross section image of the accumulative space are analyzed to extract those corresponding circular objects. Geometric parameters of the extracted connected components are scaled up to obtain the desired center coordinates and sizes of the circular objects. Results of experiments demonstrate that the proposed approach possesses higher performance than several existing ones.