Engineers at NVIDIA have created two new antialiasing modes?transparency adaptive supersampling and transparency adaptive multisampling. Both modes increase the quality and performance of antialiasing.
Transparency adaptive supersampling and multisampling take additional texel samples and antialiasing passes to enhance the quality of thin-lined objects such as chain link fences, trees, and vegetation. These types of objects are generally rendered on very simple polygon models (or even one polygon). The complexity of the final image (a group of branches or vegetation) comes from the texture that is mapped onto the polygon. Conventional antialiasing does not help this situation, because the edges of the vegetation or branches are actually inside the projected texture. Pixels inside a polygon are not touched by current antialiasing methods.
Transparency Adaptive Supersampling:
Transparency adaptive supersampling methods solve this problem by keying off information embedded in the alpha channel of the texture. Areas that have the key set can receive antialiasing even though they are not on the edge of a triangle. The result is a smoother, more beautiful image.
Supersampling is the slowest method for doing this, but does produce the highest image quality. With CineFX 4.0, the GeForce 7800 GTX uses an adaptive algorithm that applies supersampling only to selective parts of the image, so performance should be improved some.
Transparency Adaptive Multisampling:
For the best combination of performance and image quality, there is Transparency Adaptive Multisampling. It is not as high quality as supersampling, but it produces better performance.
Transparency adaptive multisampling also improves antialiasing quality?with even higher levels of performance because one texel sample is used to calculate surrounding subpixel values. Although transparency adaptive multisampling is not as high quality as the supersampling method, its increased efficiency balances improved image quality and high levels of performance.