GAMES101 Lecture 06 - Rasterization 2 (Antialiasing and Z-Buffering)

GAMES101_Lecture_06.pdf

I. Antialiasing - Sampling Artifacts

Signals are changing too fast (high frequency) but sampled too slowly.

• Nyquist-Shannon Sampling Theorem: If a function $x(t)$$x(t)$ contains no frequencies higher than $B$$B$ hertz, then it can be completely determined from its ordinates at a sequence of points spaced less than $\frac{1}{2B}$$\frac{1}{2B}$ seconds apart.

  • Higher frequencies need faster sampling.

• Aliases: Two frequencies that are indistinguishable at a given sampling rate are called "aliases".

• Convolution Theorem: Convolution in the spatial domain is equal to multiplication in the frequency domain, and vice versa.

• Sampling Artifact: Caused by sparse sampling.

Antialiasing

• Increase sampling rate:

  • Higher res. displays, sensors, framebuffers...

  • Costly and may need very high solution

• Antialiasing

  Make Fourier contents "narrower" before repeating (low-pass filter).

  • MSAA - Multi-Sample Anti-Aliasing: Approximate the effect of the 1-pixel box filter by sampling multiple locations within a pixel and averaging their values.

  • FXAA - Fast Approximate AA

  • TAA - Temporal AA

• Super Resolution / Super Sampling

  • DLSS - Deep Learning Super Sampling

II. Z-Buffering

Store current min $z$$z$ for each sample (pixel).

• Transparent objects cannot be directly processed .