A 3D imaging system is proposed in which an object is successively illuminated in at least three directions and at least three images of the object are captured by one or more energy sensors. A set of images is produced computationally showing the object from multiple viewpoints, and illuminated in the at least three directions simultaneously. This set of images is used stereoscopically to form an initial 3D model of the object. Variations in the brightness of the object provides features useful in the stereoscopy. The initial model is refined using photometric data obtained from images in which the object is illuminated in the at least three directions successively.