The parallel operator and a new circle detection method based on it are proposed in this paper which will be applied to locate circles in an image. When point A and point B are parallel pair points, the distance from A to the orientation line of B equates to that from B to the orientation line of A, and still equates to that between A and B, which is based on parallel operator. The circle detection method is constructed by the following four steps: (1) For the image, parallel operator is applied to obtain parallel pair points. (2) Compute the midpoints of the two pair points, get the midpoints energy map and determine the local maxima for locating the potential centers. (3) For each candidate center, compute the radii histogram, and whose value satisfies the threshold restriction is considered as potential radius. (4) For each center and corresponding radius, the number of points on the circle edge and the radius satisfies the threshold restriction is verified as true circle. Synthetic images and natural images are used to test the capability of the proposed method. The experimental results indicate that parallel operator has excellent detection accuracy and noise resistance, and the circle detection method has excellent performance for the detection of single circle, multiple circles, concentric circles and partial circles, and also good accuracy despite the presence of different noises.