Utilizing TAA

Unlike the previously discussed anti-aliasing methods, which only consider spatial information, TAA is based on temporal information – that is, it utilizes both the current and previous frames to smooth out these aliasing artifacts. The reason aliasing artifacts happens is because of insufficient samples; TAA solves this by sampling data over the frame sequence, significantly reducing the pressure on a single frame.

The basic idea is to apply subpixel jittering – that is, slightly shift the projection matrix of the camera for each new frame. This results in slightly different viewpoints for each frame, giving us more information about the scene than a static viewpoint would. When sampling textures during rendering, the resulting color value can be different due to the jitter. This creates a different aliasing pattern per frame, which, when accumulated over time, averages out and reduces the visible aliasing in the scene. This is demonstrated in the following...