We present a view-dependent level-of-detail algorithm for triangle meshes with subdivision connectivity. The algorithm is more suitable for textured meshes of arbitrary topology than existing progressive mesh-based schemes. It begins with a wavelet decomposition of the mesh, and, per frame, finds a partial sum of wavelets necessary for high-quality renderings from that frame's viewpoint. We present a screen-space error metric that measures both geometric and texture deviation and tends to outperform prior error metrics developed for progressive meshes. In addition, wavelets that lie outside the view frustum or in backfacing areas are eliminated. The algorithm takes advantage of frame-to-frame coherence for improved performance and supports geomorphs for smooth transitions between levels of detail. CR Categories: I.3.2. [Computer Graphics]: Picture/Image Generation
Daniel I. Azuma, Daniel N. Wood, Brian Curless, To