TAPESTRY: The Art of Representation and Abstraction
About Digital Images: Aliasing and Anti-aliasing
"Aliasing"?
The term derives from the field of "signal processing". The term aliasing describes a phenomenon related to measuring recurrent events like radio signals or sound. Your coo-coo clock may have a bird which pops out every hour on the hour, but if you pay attention (called "sampling") every 45 minutes, you might think it pops out only once every 3 hours. The resulting ambiguity or difference between "truth" and "appearance" is undesirable and is usually handled through some form of reconstructed "in-between" values--a process called anti-aliasing.
ANTI-ALIASING
In a raster image, aliasing, also called "the jaggies" or "stair-stepping", is often caused by rasterization of geometric primitives (as when you draw a diagonal line in a CAD program). It is reduced through a blending or smoothing step that softens or slightly blurs the edges of the shape.
The figure at right shows an enlarged view of a screen area through which the edge of a polygon passes at a diagonal. The left side of the image has been "anti-aliased", while the right side has not.
There are two primary approaches to anti-aliasing, super-sampling and filtering. Both are illustrated on the Examples page. Below is a description of each process in the context of displaying the end of a solid-colored polygon.
Super-Sampling
The super-sampling approach creates a temporary high-resolution grid where the edge is being drawn, rasterizes the edge at that higher resolution, then sets the actual pixel colors based on the number of sub-pixels colored by the edge using a blend from white (the background) to the edge color.
Super-sampling can also be used by raster rendering processes like ray-tracing by sampling subpixels of each pixel to develop greater nuance in the image.
Filtering
The filtering approach rasterizes the image at normal display resolution, then smooths the image's edges with image-processing techniques that examine a pixel's neighbors to blend a new color. This is generally faster, but degrages the image more than super-sampling.
This procedure is computationally similar to the Photoshop "smooth" function.
Benefits
At any spatial resolution an anti-aliased image will generally look "better" than one which has not been anti-aliased.
Costs
As ever, the primary cost is TIME. A second cost is a blurring at the edge of the object which might be seen as a reduction in object definition. Finally, anti-aliasing blends foreground and background; if you need to overlay the anti-aliased image on another background, an edge anti-aliased between green and white can be an irritation rather than a plus if the new background is dark.
Last updated: April, 2014
