QuickStudy: Anti-aliasing

Computerworld - Anti-aliasing smooths the raw and haggard edges on digital type and images on computer and handheld displays, wireless phones, printers, even digital cameras.

Aliasing - jagged or stair-stepped edges - appears when there are too few pixels in an image or on a display to represent it realistically.

Spectral aliasing - moire or herringbone patterns where none should be - appears as a result of interference between digital signals, such as color and brightness.

How It Works

In graphics or type, anti-aliasing algorithms sample, or examine and evaluate, the colors and shades of pixels adjoining curved or diagonal lines. They also shade some pixels to create a softer line. The eye no longer clearly sees the stair-stepped edge and, paradoxically, it reads the softer line as clearer.

Dithering, or creating a similar shaded effect using patterns of dots of colors, can create an anti-aliasing effect for photographs and shaded drawings.

Increasing the resolution and thus reducing the pixel size can create an effect similar to that of anti-aliasing, but this may not always be feasible.

Fractal compression lets you store thousands of large image files on a single CD.

Fractal compression works by reducing each shape or pattern to an equation. When the fractal image is reconstructed, there's often more information than the display can show, and artifacts - bits of white debris - are created. Anti-aliasing eliminates these artifacts and creates smoother edges.

High-end graphics hardware maker Silicon Graphics Inc. in Mountain View, Calif., developed an anti-aliasing method that uses special accumulation buffers that temporarily store rendered frames. When several frames have accumulated, the graphics chip blends them together.

To counter aliasing, some devices, such as Intel Corp.'s Pentium III chip and many printers and digital cameras, incorporate anti-aliasing filters. High-end graphics cards may contain more sophisticated anti-aliasing filters.

Most anti-aliasing filters sample adjoining pixels between two and 16 times at different locations, then combine and average the different readings to get the most realistic color. More samples result in subtler gradations between image edges and along curved or diagonal lines and type. But a higher sample rate takes more time and memory.

Some cards use supersampling, which renders the image at a resolution higher than the display can support and then scales down and filters the image - in effect, adding new pixels, before sending it to the display. But such brute-force techniques take a toll on performance.