Abstract : This study presents a video-based experiment for the study of markerless human motion. Silhouettes are extracted from a multi-camera video system to reconstruct a 3D mesh for each frame using a reconstruction method based on visual hull. For comparison with traditional motion analysis results, we set up an experiment integrating video recordings from 8 video cameras and a marker-based motion capture system (Vicon™). Our preliminary data provided distances between the 3D trajectories from the Vicon system and the 3D mesh extracted from the video cameras. In the long term, the main ambition of this method is to provide measurement of skeleton motion for human motion analyses while eliminating markers.