This paper describes the filtering used in Microsoft ClearType. ClearType is a software system that enhances the resolution and readability of fonts on displays that contain a repeating pattern of addressable colored sub-pixels. The filtering in ClearType is based on a perceptual model of human vision. The perceptual model leads to an optimization technique for finding the best output values. The results of the optimization can be approximated by pre-filtering each color channel of an input image and then sampling each filtered color image at the spatial locations of the same colored sub-pixels in the display. We refer to this filtering followed by displaced sampling as RGB decimation. RGB decimation eliminates the phase error caused by standard anti-aliasing. A further approximation of the optimal filter yields RGB decimation with displaced box filters. Fourier analysis demonstrates that both the optimization technique and the displaced box filter suppress frequencies that contribute most to color fringing in unfiltered displaced sampling.