This paper presents a vision-based Leader-Follower cooperative scheme which consists of two underwater vehicles in the absence of explicit communications and direct information interchange. A novel method for implicitly calculating the relative pose between the two vehicles is introduced. The absolute and relative localization algorithm is solely based on the observation of visual features projected on the vehicles common workspace and it is performed in a strictly decentralized manner. The cooperation task consists of a Leader vehicle which is tele-operated in an open-loop fashion while inspecting a flat surface and a Follower vehicle which follows the Leader while keeping a fixed 2D distance offset. The Follower is able to track the Leader's motion at all desired configurations via a motion tracking controller designed accordingly. The proposed control scheme for the Follower is relayed on vision-based implicit communications data and has analytically guaranteed stability and convergence properties. The accuracy and reliability of the implicit vision-based localization system, the performance of the designed control scheme, as well as the efficiency of the overall system in the proposed cooperative task, are experimentally verified using two small Remotely Operated Vehicles (ROVs) in a test tank.