3d Digitization of Heritage: Photometric Stereo Can Help

Abstract

Abstract. There are mainly two families of photographic 3D reconstruction. Photogrammetry techniques work according to the principle of triangulation, from the matching of different views, while photometric techniques link the appearance of a 3D point to the orientation of its normal, relative to the direction of the incident light. While photogrammetry allows to find the global shape of a 3D scene, if it is sufficiently textured, photometric techniques highlight the details of the relief, as long as the model linking the lighting to the shape and reflectance of the scene is sufficiently realistic. In order to avoid errors in the 3D models obtained, all the photographic techniques of 3D reconstruction have benefited, over the years, from algorithmic improvements that make them more and more robust to outliers or unreliable data. Moreover, the complementarity between these two types of approaches having been identified for a long time, many solutions have been proposed to merge them. Our work aims at providing the free and open-source photogrammetry software Meshroom with the benefits it could get from photometric stereo, particularly in the context of the 3D digitization of heritage, knowing that it is the only photometric technique for 3D reconstruction that has really proven itself.

3D surface Approximation of the Entire Bayeux Tapestry for Improved Pedagogical Access

“…The photometric stereo technique (PS) [42] provides a solution to this high-frequency 3D-reconstruction problem. Furthermore, it has successfully been applied in various cultural heritage applications [17,31]. Therein, a per-pixel surface normal map is estimated from a series of photos taken under the same viewing angle, while varying the lighting directions to create shading effects in the images.…”

Neural detection of spheres in images for lighting calibration

“…For example, the 3D digitization of heritage allows to preserve the object of study while simplifying its access to the general public, as well as to professionals who, by having a digital copy faithful to the original, will be able to use it for their analyses. 2 Photogrammetric methods, which are well known, provide a surface reconstruction from the correspondence between images taken from different viewpoints. However, photogrammetry does not explicitly use the photometric characteristics of the scene.…”

“…To do this, it is customary to position a sphere in the scene, which is visible in each of the images. 2 If the 3D reconstruction method requires changing the camera pose, or if the sphere must be held at the end of a pole when access to the scene is limited, this implies that the position of the sphere varies from one image to another, as illustrated in the example of Figure 1. Manual trimming of the sphere in each image then becomes very tedious.…”