Interactive Media Systems, TU Vienna

Light Source Estimation with RGB-D Camera

Master Thesis

About This Topic

The estimation of environmental illumination in a real scene is an important problem addressed in many fields of computer graphics, particularly in augmented reality, photorealistic rendering, appearance capturing and the others.

The main goal of this thesis will be to develop an algorithm for estimation of spatially-varying illumination from a real environment. The RGB-D camera will be utilized to capture the real geometry which will assist in the estimation of illumination. Additionally, the shadows will be detected in the RGB image to calculate the positions and intensities of light sources. By moving the RGB-D camera throughout the real environment the estimated illumination can vary in the spatial domain. Spatial variability and temporal coherence of estimated light will be achieved by accumulating information from multiple frames.