| With the advances in large scale model acquisition technology the complexity of geometric models has been increasing dramatically over the past decades. Processing gigantic models consisting of hundreds of millions of triangles has become a requirement in many computer graphics applications. The progress in the performance of CPUs and graphics processing units (CPUs) has been orders of magnitude better than the improvement in the fast hard disk technology during the same period, and this gap is widening everyday. Hence the performance bottleneck of graphics processing and rendering algorithms has been pushed from rendering to disk-to-memory data transfer rates today.;In this dissertation, we first proposed a unified framework to compress different attributes of a generic point cloud such that the complexity of the data set describing the corresponding 3D objects is reduced and thus the frequency of I/O operations at runtime. In the proposed scheme, the coding process is led by an iterative octree cell subdivision of the object space. We employ attribute-dependent encoding techniques to exploit different characteristics of various attributes. All of these have led to significant improvement in the rate-distortion (R-D) performance and a computational advantage over the state of the art. Furthermore, given sufficient levels of octree expansion, normal space partitioning and resolution of color quantization, the proposed point cloud encoder can be potentially used for lossless coding of 3D point clouds.;Secondly, we explore the impacts of primitive attributes, including position, normal and connectivity, on the data access patterns and hence the performance of general 3D model processing algorithms. We demonstrate that the performance of general 3D model processing algorithms can be improved by simply reorganizing primitives based on the relative importance of different primitive attributes as used in an algorithm. In this dissertation, we first present a novel data layout algorithm that generates data layouts based on the relative importance of different primitive attributes designated by users without any requirement of specific runtime data access pattern. Next, we testify the benefits of our layout algorithm through typical 3D model processing applications including view-dependent rendering, feature sensitive clustering and connectivity-driven primitive traversing. |
