The term ‘aliasing’ refers to how integer multiples of frequencies can be confused for one another under certain sampling assumptions.
For example, imagine a clock at one. I ask you to close your eyes and then I rotate the clock fully around back to one before I ask you to open your eyes. From your perception, how many rotations occurred in the time your eyes were closed?
From the information you have, it could be 0, 1, 2, … , etc. Any integer number of rotations would match the information you have – that the clock returned to its original position in the time your eyes were closed.
These are ‘aliases’ of one another – they all refer to same information you possess just like all aliases for a single person refer to that single person.
The above is what is known as ‘temporal aliasing’. When you’re talking about your computer graphics you get a similar phenomenon known as ‘spatial aliasing’.
Your monitor (and video signal) have a certain resolution. However, the image they’re trying to display is philosophically analog – you’re trying draw circles, diagonal lines and other elements that do not exactly match up against the pixel sampling of your monitor.
For large-scale images, this isn’t a big deal. But as you get into smaller elements of those images, you end up with the pixels being ‘too thick’ to accurately represent what you want to see. In essence, you’re attempting to reproduce high frequency signals on a low frequency device and you’re inaccurately ‘aliasing’ them to those lower frequencies.
In temporal aliasing, you normally solve the problem by filtering out the high frequencies. In spatial aliasing, those high frequencies are actually what you’re interesting in, so you normally solve the problem by blending nearby frequencies.
TL;DR: Anti-aliasing in graphics smooths other otherwise blocky or jagged portions of the graphics.
Latest Answers