This paper presents a novel approach to human gait analysis using a marker-free system. The devised acquisition system is composed of three synchronized and calibrated charge coupled device cameras. The aim of this work is to recognize in gray level image sequences the leg of a walking human and to reconstruct it in the three-dimensional space. An articulated threedimensional (3D) model of the human body, based on the use of tapered superquadric curves, is first introduced. A motion-based segmentation, using morphological operators, is then applied to the image sequences in order to extract the boundaries of the leg in motion. A reconstruction process, based on the use of a least median of squares regression is next performed, to determine the location of the human body in the 3D space. Finally, a spatial coherence is imposed on the reconstructed curves in order to better fit the anatomy of the leg and to take into account the articulated model. Each stage of the proposed methodology has been tested both on synthetic images and on real world images of walking humans. The obtained results suggest that this approach is quite promising and should be useful in the study of the gait.