In the previous studies, basic methods to estimate motion or attitude of the vehicle (camera), i. e. ego motion, from video image data obtained continuously by a VTR are proposed, and their performances and applicability to the actual images are investigated. By applying an iterative estimating algorithm based on the Extended Kalman Filter to continuous pictures of the ground surface, it is assured that the motion can be estimated with enough stability. In this study, continuous underwater images are obtained by a ROV, and its motion is estimated by the same algorithm to investigate applicability of the methods for actual use in the water. The accuracy of the estimation is evaluated for various conditions, and that of 0.25 edgree (maximum error) in yaw angle is assured for ordinary images with appropriate landmarks. The necessity of calibration for the edge distortion is pointed out in case very few landmarks exist in the pictures.