Splat/Mesh Blending, Perspective Rasterization and Transparency for Point-Based Rendering

Abstract

In this paper we present multiple simple and efficient improvements for splatting based rendering systems. In a first step we derive a perspectively correct splat rasterization algorithm suitable for both efficient implementation on current GPU and the design of fast dedicated rasterization units taking advantages of incremental calculations. Next, we propose a new efficient and high-quality approximation of the optimal EWA (Elliptical Weighted Average) filtering framework. We also show how transparent point-clouds can be rendered by current GPU in an order independent way by using a modified depth-peeling approach. Finally, in the context of hybrid points and polygons rendering systems, we present a simple mechanism to smooth the transitions between the different representations. Combined to previous contributions in the field, these improvements lead to a high-quality, performant and fullfeatured hardware-oriented splatting rendering system.