This paper describes the implementation of a robotic system using artificial vision as a computational tool developed for the evaluation of kinematic parameters during underwater gait therapy. This work is focused in the estimation of some angular kinematic variables during the gait process through video processing techniques. Two video acquisition protocols that differ in the limitations imposed to patients are proposed, these protocols lead to different processing methods. This study is entitled in a research project aimed at developing a tool that allows evaluating the underwater therapy benefits as part of the lower members' rehabilitation in patients that present human gait pathologies.