Arbitrarily shaped objects relighting using an RGB-D camera

Usually, relighting techniques require knowledge about the shape of the target object and the lighting environment. The quality of the result is highly dependent on the normals of the object because they are used in the computation of the illumination. In this paper, we propose a new relighting approach for arbitrarily shaped objects using an RGB-D camera such as the Microsoft's Kinect. The depth map is useful to estimate the normals of the object, but can be inaccurate because of the noise such as discrete depth values or missing data. An accurate segmentation of the target region for relighting is also an open issue since the boundaries in the depth map does not always match color's ones. We focus on the depth map modification to segment the object region and normal estimation for accurate relighting. In our experiments, we adapted some normal estimation methods from modified depth map and evaluated the accuracy of the relighting results. We discuss the effectiveness of relighting approach for an arbitrarily shaped object and the possibility of a real time relighting.