Mixture of Spherical Distributions for Single-View Relighting
IEEE Transactions on Pattern Analysis and Machine Intelligence
Proceedings of the 5th International Conference on Ubiquitous Information Management and Communication
Hi-index | 0.00 |
Several techniques have been developed for recovering reflectanceproperties of real surfaces under unknown illumination conditions.However, in most cases, those techniques assume that the lightsources are located at inifinity,which cannot be applied to, forexample, photometric modeling of indoor environments. In thispaper, we propose two methods to estimate the surface reflectanceproperty of an object, as well as the position of a light sourcefrom a single image without the distant illumination assumption.Given a color image of an object with specular reflection as aninput, the first method estimates the light source position byfitting to the Lambertian diffuse component, while separating thespecular and diffuse components by using an iterative relaxationscheme. Moreover, we extend the above method by using a singlespecular image as an input, thus removing its constraints on thediffuse reflectance property and the number of light sources. Thismethod simultaneously recovers the reflectance properties and thelight source positions by optimizing the linearity of alog-transformed Torrance-Sparrow model. By estimating the object'sreflectance property and the light source position, we can freelygenerate synthetic images of the target object under arbitrarysource directions and source-surface distances.