Lecture 7: Intro to Geometry, Splines, and Bezier Curves (13)
GuardHei
One recently popular representation in video games is signed distance field, which is a form of distance functions.
It is widely used to approximate the polygon mesh and naturally comes with rounded edges. Many games utilize SDF to render alias-free fonts, approximate far-distant shadows (e.g. unreal 4), simulate a coarse software conetrace (e.g. unreal 5's mid-range global illumination solution) and speed up volumetric cloud rendering (especially some high-end mobile games).
wangjim19
One interesting tradeoff between the different ways to represent geometries is the tradeoff between space complexity and efficiency of obtaining certain information. For example, if we store a 3d array of points where each element is 0 if unoccupied, or 1 if occupied, then it becomes very efficient to query whether a point is occupied, but it is space inefficient to store.
xinwei-zhuang
Prof. mentioned implicit surface is better and more efficient for ray-tracing, but not sure why's that..
Staffrishiu
@xinwei-zhuang thats a great question - the key reason is because the computation heavy part of ray-tracing is in intersecting the rays with the geometry and this is way easier for implicit surfaces
One recently popular representation in video games is signed distance field, which is a form of distance functions.
It is widely used to approximate the polygon mesh and naturally comes with rounded edges. Many games utilize SDF to render alias-free fonts, approximate far-distant shadows (e.g. unreal 4), simulate a coarse software conetrace (e.g. unreal 5's mid-range global illumination solution) and speed up volumetric cloud rendering (especially some high-end mobile games).
One interesting tradeoff between the different ways to represent geometries is the tradeoff between space complexity and efficiency of obtaining certain information. For example, if we store a 3d array of points where each element is 0 if unoccupied, or 1 if occupied, then it becomes very efficient to query whether a point is occupied, but it is space inefficient to store.
Prof. mentioned implicit surface is better and more efficient for ray-tracing, but not sure why's that..
@xinwei-zhuang thats a great question - the key reason is because the computation heavy part of ray-tracing is in intersecting the rays with the geometry and this is way easier for implicit surfaces