What is difference between A buffer algorithm Z buffer algorithm?

Answer 1

A z-buffer is a raster buffer that stores color and depth information at each pixel. The "z" in the title refers to the "z" plane in 3D space, which is traditionally thought of as the "depth" dimension.

The buffer initializes each pixel to the default color and an infinite depth. During the rendering process, when a color is written to a pixel, it first compares the current depth of the color in the pixel. If the new color is closer than the current color but closer than the clip plane (which is typically zero), the color is written and the depth updated.

In that sense, it's similar to the painter's algorithm, where the closer object covers the further object.

Here's the basic algorithm:

WritePixel(int x, int y, float z, color c)

if ( z < zbuffer[x][y] && z > 0 ) then

zbuffer[x][y] = z;

frameBuffer[x][y] = color;

end

The a-buffer uses the same algorithm for handling depth, but adds anti-aliasing. Each pixel contains a set of sub-pixels. During the write operation, the values are accumulated at the sub-pixel level. For the final pixel read, the final color is the sum of all the sub-pixels.

The algorithm was originally developed by Loren Carpenter (or Pixar) for the RenderMan renderer. The position of the sub-pixels in each pixel are randomly selected in space and time, which allows smooth blurring of moving objects. RenderMan dices geometry down to micropolygons (polygons approximately the size of a pixel), and then performs a coverage test to determine if a sub-pixel is covered by a micro-polygon.

However, this approach doesn't work with a more "typical" renderer, since they typically deal with points, which unlike micropolygons, have no surface area.

A common adaption of this algorithm is the accumulationtechnique, which renders an image multiple times, randomly jittering (moving) the position of the eyepoint by some small amount. The result of each rendering is accumulated and averaged into single buffer. This approach is made practical with a hardware accelerated renderer such as OpenGL. However, this approach is probably better thought of as supersampling rather than an a-buffer.