Rendering Surfaces with Bidirectional Texture Function by Radiosity Method

Surfaces with bidirectional texture function(BTF)is rarely rendered by radiosity method,because their materials are normally defined by point based sampling,which does not adapt to the conventional patch-based process of radiosity approach.An efficient algorithm for integrating BTF surfaces into the framework of radiosity is presented.For a BTF sample set,a clustering process is first applied to all pixels according to their intensity divergences.Then in each pixel cluster,all the viewing directions with similar shading effects are clustered further.Each refined cluster is fitted by a low frequency illumination model,and the high frequency deviation of the estimated shading from the real intensity at each pixel within the cluster regarding each viewing direction is calculated.The low frequency model is assigned to the local surface region as its average reflectance for radiosity calculation.The high frequency model is employed for building the detail on the surface.Our approach achieves high compression rate and generates interrefection shading effects on the BTF surfaces.GPU is also used to speed up the rendering performance.