3D Object Reconstruction using Computer Vision

This book belongs to the Computer Vision domain, more specifically, to the area of Three-Dimensional (3D) Vision. The 3D information of the surrounding environment perceived by human eyes is one of the most important kinds of information offered by sight. Thus, a large effort has been developed by the scientific community throughout time in the development of methods that make possible the acquisition of a scene (or object) using techniques of Computer Vision. Computational methods were developed to obtain 3D models of the human body from image sequences. On both the image pixel information is transformed into 3D spatial coordinates. The images are acquired by moving the camera or the object to be reconstructed, without imposing severe restrictions, on either the movement involved or on the illumination conditions. The first method adopts the Stereo Vision approach, in which a depth map of scene objects is recovered. The second method is based on a volumetric approach for Shape-from-Silhouettes, which constructs a 3D shape estimate (Visual Hull) of an object. The methods were assessed using objects with dissimilar textures, shapes and sizes, including real human body structures.