Reectance Transformation Imaging is a technique that provides a digital and useful representation of an objectthrough photometric and geometric local assessment of the surface. RTI technique consists in acquiring asequence of images from a xed observation position while varying the direction of the light source aroundthe observed object. Thanks to a further reconstruction process, the continuous angular reectance for eachpixel can be computed from the set of discrete acquisitions and rendered interactively. Currently, the mostused mathematical functions that allow this reconstruction from RTI's acquisitions are Polynomial TextureMapping (PTM), a method based on Hemispherical Harmonics (HSH) and most recently the Discrete ModalDecomposition (DMD). For these three approaches, a uniform spatial distribution of light sources is an implicithypothesis. In practice, it is often not possible to achieve this uniform spatial distribution due to intrinsiclimitations in systems or in the acquisition conditions. It is then necessary to take into account this non-uniformity in order to avoid artifacts that could alter modelling and subsequent visual rendering. To addressthis issue, we propose a methodology consisting in the estimation of the local density of the lighting directionsused during RTI acquisition. These values are then used to generate a weight for each light position enabling tocorrect its contribution in the regression performed during the tting.
展开▼
