Multiresolution model generation of texture-geometry for the real-time rendering

Abstract

The multiresolution model is used to satisfy requirements for the quality of a rendered scene and rendering time. In this paper, a new method for constructing a multiresolution model is proposed. The multiresolution model represents not only polygonal geometries but also texture images. The multiresolution geometry should preserve geometric features for an object. These features are object-centered features, camera-centered feature, and light-centered features. M-AGSphere structure is proposed to process camera-centered and light-centered features in real-time. It has an efficient structure to manipulate complex polygon in real-time. The runtime cost is compared with the popular method using an edge collapsing operator. The multiresolution geometry has problems of texture mapping. These problems are identified as the boundary problem and the distortion problem. The boundary problem is solved by generating a texture image to preserve continuity of the geometry. Furthermore, the new texture map is generated by minimizing distortions. These distortions are identified by the multiresolution mapping distortions which consist of the pixel resolution distortion and the polygon resolution distortion. The proposed method minimizes both the pixel resolution distortion and the polygon resolution distortion. Starting from the texture map which is generated for the multiresolution geometry, we propose an algorithm to hierarchically enhance the texture map. The texture map is enhanced by sub-maps with more detail and less distortion. We also identify a set of geometry that has a valid texture mapping function over a newly enhanced texture map. Along with the multiresolution geometry, these hierarchically enhanced textures consist of the multiresolution model of the texture-geometry.