Short Paper: Approximating Distance Fields in Image Space

Abstract

Fast proximity query algorithms are needed for a vast majority of computer graphics applications, ranging from visualization, simulation, animation, modelling, virtual and augmented reality to computer games. We present a novel and fast approach to approximate distance fields in image space at interactive frame rates. The vicinity of a point is sampled and the minimal distance to its surrounding geometry is determined based on the scenes depth values obtained by multiple stencil routed A-Buffers. Two sampling methods are compared with ground truth data in image space precision. Additionally, we introduce a method to determine the distance of an object to the next occluded object and a method to sample a directed distance field. The algorithms perform entirely on the GPU at interactive frame rates without the need for any precomputations and can handle dynamic content of arbitrary and deformable objects in massive data sets, making them feasible for a variety of applications.