High-performance architecture for anisotropic filtering

Anisotropic filtering has been developed as a way of increasing the quality of texture mapping. However, a real hardware implementation of anisotropic filters implies approximations and simplifications that result in lower quality. In this paper we present a new, efficient hardware oriented anisotropic filtering technique. Specifically, the new algorithm we propose is based on the utilisation of a new distance computation method. This distance generation scheme permits simplification of the algorithm without reducing its quality. Additionally, a classification procedure for the identification of the position of each texel inside the footprint is proposed. This simple procedure permits the reduction of the computational requirements associated with the algorithm. We also present a new method for the computation of the coverage of a texel, based on the storage of a reduced set of possible coverage patterns. The new distance scheme obtains the patterns of all texels from a given texel through simple shift operations. The good quality results obtained with the filtering algorithm we propose, together with the low computational and storage requirements of the architecture we present, makes it a good candidate for hardware implementation.