Frameless rendering

Abstract

This thesis studies a render paradigm of computer graphics called "frameless rendering". Breaking with the traditional approach in computer graphics, it allows an image to be displayed even when it has not been finished yet. It ensures that the user is constantly provided with the possibility to immediately see the results of his interactions on screen - even if this means to see only partially up-to-date imagery. The idea differs from classic single-buffered approaches in two aspects. First, the image is not rendered from top to bottom; rather the next to-be-computed pixel is always chosen at random. Second, the approach allows to display the currently available image at every point in time. In the most strict consequence this means that there really are no coherent images anymore, no connected image sequences, no frames per second, no frames at all - hence the name "frameless rendering". Frameless rendering has evolved to an alternative render paradigm, moving from a camera-based to a perception-based way to understand the image generation process in computer graphics. However, only very few researchers have dedicated their work to this idea since its introduction. No "theory" of frameless rendering has been developed ever since and only a couple of researchers have tried to formalize the process in some way. We fill this gap by providing an in-depth exploration of this novel approach to digital image synthesis. We examine the effectiveness of the traditional random pattern as a sampling order during frameless rendering and analyze the suitability of alternative deterministic approaches. In addition, we reformulate the rendering equation to account for the frameless rendering approach and show that by finding a suitable reconstruction filter, frameless rendering can deliver better imagery than traditional rendering approaches in scenarios of a very limited computational budget.