Week 10 - Triangularization and Rasterization
0Today we talked about two topics - given an input set of points, how to create a set of triangles from it. The braindead way is to just sort them from left to right, but this is going to create weird slivers of triangles, so a better way is to enforce the Delaunay criteria on it, which is to say that given any four points in a quadrilateral, it is better to split it so that the edge going down the middle is minimized. Another way of saying it is that given any two points in a triangularization, there are no other points within the radius of the circle formed by the two points. (In other words, there's no other points closer.)
Second, we talked about rasterization, which is the process by which lines, triangles, quads, etc., get "printed" on the screen in the form of pixels. Doing it poorly is easy - just go through every pixel on the screen and see if its center point is within a triangle, and if it is, color it in. Save the depth in a z-buffer so that if you try drawing over it with another triangle, the closer one will be drawn. But this will give bad performance with lots of triangles, so there are a lot of very clever techniques to cull the triangles we need to process.