VSD factors, classes, Z/A-buffer

Factors in choosing a visible-surface algorithm

1. Scene complexity: number of polygons and surfaces (a few objects or thousands).
2. Type of objects: closed convex solids, concave objects, curved surfaces, transparent objects.
3. Memory available: the Z-buffer needs a full depth buffer; the A-buffer needs even more.
4. Speed required: real-time (games) or offline rendering.
5. Hardware support: many GPUs implement the Z-buffer in hardware.
6. Image quality: need for anti-aliasing, transparency, resolution independence.
7. Ease of implementation: simple algorithms are preferred when quality needs are low.

Two classes of visible-surface detection methods

1. Object-space methods: compare objects and parts of objects with each other in the 3D scene to find which surfaces are visible. Examples: back-face detection, depth sorting (painter's), BSP tree. Precision is that of the object description and the result is independent of screen resolution, but the cost grows quickly with the number of objects.
2. Image-space methods: visibility is decided at each pixel position of the projection plane by finding the nearest surface. Examples: Z-buffer, A-buffer, scan-line, ray casting. Accuracy is limited by resolution; they are simple and efficient for complex scenes.

Limitation of Z-buffer

Only the nearest surface per pixel is stored, so transparent surfaces cannot be shown, and edges are aliased.

How the A-buffer overcomes it

The A-buffer keeps a linked list of surface records at each pixel position (each with RGB, opacity, depth, area coverage, surface id and pointer). Depth field ≥ 0 means a single surface; depth field < 0 means the intensity field points to the list. After all polygons are processed, the surfaces at each pixel are sorted by depth and blended using their opacity and coverage, giving transparency and anti-aliased edges. It needs more memory and time.