A graphics processing system in which sub-pixel correction is implemented in a new and more economical way. A half-pixel offset is originally imposed on both the X and Y axes, so the sub-pixel correction value may be positive or negative. A relatively coarse resolution is used for subpixel correction, so that the DDAs (and other functional blocks) do not have to perform a full multiply: instead they merely perform simple add operations (addition of partial products) to derive the necessary offset from the delta-X values, using a proportionality constant provided by the rasterizer. The DDAs preferably include a "guard band" in their calculations, so that values which exceed the maximum (e.g. 255, for 8 bits of subpixel resolution) do not wrap to zero; instead the output is held at the maximum until the computed value comes down below the maximum.