Edge function based anisotropic texture filtering

Abstract

Texture mapping is a technique that most effectively improves the realism of computer-generated scenes in 3D Graphics. Tri-linear filtering of the mipmapped texture has been popular as a texture filtering method but it blurs images on the surface of objects angled obliquely away from the viewer in a scene. Various anisotropic filtering methods like footprint assembly, Feline, and fast footprint mip-mapping have been proposed to satisfy the desire for the high quality image. In spite of increases to memory bandwidth, the limit of the memory bandwidth is still a bottleneck of the texture filtering hardware. Moreover, it is very important to keep the quality of rendered image good. In this thesis, we propose an edge function based anisotropic texture filtering The weight plays a key role in effective filtering to render the image of high quality with the restricted number of texels loaded from memory for real-time filtering. So, in generating weights that are used at post-filtering texels in a region that covers the footprint, our anisotropic filtering algorithm uses both an exact footprint shape approximation and Gaussian filter. For an exact footprint shape approximation, the edge function is used to represent the edge of the footprint. The normalized value of the edge function gives the distance relative to the contribution of texels to a final intensity. Calculating a Gaussian filter using this normalized value generates a good weight. Therefore, the quality of rendered images is superior to other anisotropic filtering methods in the same restricted number of texels. For images of the same quality, our method requires less than 0.625 times the texels of other methods. Consequently, the improvement in performance is more than twice that of other methods. With low hardware overheads, our method can be implemented at a reasonable cost. In practice, the algorithm is demonstrated through VLSI implementation. The hardware, which is described by verilog and synthesiz...