Pareto-Based Perceptual Metric for Imperceptible Simplification on Mobile Displays

Abstract

Graphics on mobile devices has become popular because untethered computing is convenient and increases pro- ductivity. Mobile displays come in different resolutions that affect the scene Level-of-Detail (LoD) that users can perceive: smaller displays show less detail, making lower resolution meshes and textures acceptable. Mobile de- vices frequently have limited battery energy, low memory and disk space. To minimize wasting system resources, we try to render mobile graphics scenes at the lowest LoD at which users do not perceive distortion due to sim- plification. We call this LoD the Point of Imperceptibility (PoI). The PoI LoD depends on several factors including screen size, scene geometry and lighting levels. We propose a perceptual metric that identifies the PoI LoD of a target mobile display and accounts for object geometry, lighting and shading. Our perceptual metric generates a screen-dependent pareto distribution with a knee point that corresponds to the PoI. We employ wavelets for simplification, which gives direct access to the mesh undulation frequency that we then use to parametrize the perceptual CSF curve.