To analyze the object's motion from images, the velocity and the motion parameters (scale of enlargement or shrinkage, angular velocity) must be required. The velocity in images can be obtained from the optical flow equation. However, it is not reasonable to calculate such motion parameters from the velocity estimated on the assumption that it translates. Therefore the motion parameters based on the optical flow may have systematic error. We propose a new algorithm that can estimate both optical flow and the motion parameters at the same time after scanning images only once.