LOD selection driven by per-level approximation errors: the splat budget is spent where it
removes the most error per splat, using the manifest's error tables when present and errors
derived from splat counts otherwise. Lifts sparse, low-quality regions that distance alone
leaves coarse, at the cost of holding noticeably more source data in memory.
LOD selection driven by per-level approximation errors: the splat budget is spent where it removes the most error per splat, using the manifest's error tables when present and errors derived from splat counts otherwise. Lifts sparse, low-quality regions that distance alone leaves coarse, at the cost of holding noticeably more source data in memory.