A dense network of strong motion measurement is significant for the earthquake hazard mitigation. To realize such a network, this paper proposes a strong motion measurement using video images recorded in security video cameras. Even though the accuracy is sacrificed to some extent, this measurement enables us to use video cameras as maintenance-free and simple apparatus which are installed in many buildings. The image analysis and the inversion are developed. Several tests were carried out to verify the validity of the methods, and it is shown that input acceleration can be fairly predicted from recorded video images.