Abstract
AbstractHardware tessellation is de facto the preferred mechanism to adaptively control mesh resolution with maximal performances. However, owing to its fixed and uniform pattern, leveraging tessellation for feature‐aware LOD rendering remains a challenging problem. We relax this fundamental constraint by introducing a new spatial and temporal blending mechanism of tessellation levels, which is built on top of a novel hierarchical representation of multi‐resolution meshes. This mechanism allows to finely control topological changes so that vertices can be removed or added at the most appropriate location to preserve geometric features in a continuous and artifact‐free manner. We then show how to extend edge‐collapse based decimation methods to build feature‐aware multi‐resolution meshes that match the tessellation patterns. Our approach is fully compatible with current hardware tessellators and only adds a small overhead on memory consumption and tessellation cost.
Highlights
For real-time applications, complex geometric models are usually rendered by mean of level-of-details (LOD) [LWC⇤02]: input meshes are adaptively downsampled to satisfy a view-dependent error criterion
A transition from a fine to a coarse integer factor is accomplished by gradually collapsing pairs of vertices. This fractional tessellation scheme is very effective at removing popping artifact when adaptively evaluating subdivision surface [NLMD12]
As mentioned in the introduction, a more radical approach consists in disabling fractional tessellation and implementing some kind of tri-linear interpolation of the displacements across the levels [SPM⇤13]
Summary
For real-time applications, complex geometric models are usually rendered by mean of level-of-details (LOD) [LWC⇤02]: input meshes are adaptively downsampled to satisfy a view-dependent error criterion. Each input patch (a triangle or a quad) can be subdivided at an arbitrary and continuous level [Mor01] enabling patch-grain LOD control with smooth spatial and temporal transitions (Figure 1) In this context, the input geometry is often decomposed into a coarse base mesh and a displacement map from which the positions of the vertices generated by the tessellation engine are reconstructed [TBB09]. If starting from the coarser level, it is not possible to control topologically where new vertices will be added This severely limits the ability to construct feature-aware LOD. A transition from a fine to a coarse integer factor is accomplished by gradually collapsing pairs of vertices This fractional tessellation scheme is very effective at removing popping artifact when adaptively evaluating subdivision surface [NLMD12].
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have