An enhanced filtering-based approach to approximate volumetric ambient occlusion

Abstract

Achieving, through shading, a good visualisation of scientific volumetric data sets is prohibitive for interactive applications due to a significant amount of existing occlusion effects and lighting variation. Ambient occlusion, although allowing us to understand these data through soft shadows, often comes with an important cost of pre-processing which prevents its use especially in the applications based on the transfer function editing. In this paper, we present a new method for approximating ambient occlusion for volumetric data using an exponential function which allows for high-quality hardware-accelerated filtering and leads us to achieve more compact information using a filterable representation in order to avoid the classic occlusion test with a binary manner. Our evaluation of a series of medical images demonstrates that the proposed algorithm provides smooth shading, producing high quality results, achieves an interactive rate and consumes less memory.