The current method of verification for external beam radiation therapy visually compares a simulation image with a portal image. However, because this method depends largely upon the observer's experience, it sometimes results in inter-observer differences. In this study, we developed software to measure atomatically the quantitative difference between the simulation image and portal image using an image-analysis method. The feasibility of this software was evaluated on a rectangular field in the pelvic region. We took 12 simulation images of a pelvic phantom, setting 4 different field shapes on each of 3 isocenters. We then obtained 84 portal images setting 7 known distances from each of the 12 simulation images. Using this software, the direction of shift was detected correctly, and the distance of shift was detected correctly to within less than 3 mm. We consider that this software could be a useful method of verification.