Vision-based feature extraction of a synchronized swimmer doing figures was done in the hopes of correcting and improving a synchronized swimmer’s ability to do figures. Extracted feature through image processing techniques plays an important role in understanding the synchronized swimmer’s body kinematics and dynamics. In gathering data, two cameras were used to video capture the swimmer’s movements. One was stationed at a calculated distance above the water, and the other camera was placed underwater directly above the other camera. In this article only the underwater feature extraction of the three figures synchronized swimming, Porpoise, Neptunus, and Ballerina, and the essential image processing techniques used for the feature extraction were covered. The results showed that empirically extracted threshold limits for the skin region produced smooth feature extracted foregrounds.