Algorithms and hardware for data compression in point rendering applications

Abstract

The high storage requirements associated with point rendering applications make the utilization of data compression techniques interesting. Point rendering has been proposed only recently and, consequently, no compression strategies have yet been developed. In this paper we present compression algorithms for two specific data distributions widely used in point rendering: a naive distribution with no specific sorting of the points and a layer distribution which is suitable for incremental algorithms. In this last case points are sorted in layers and the connectivity among them is encoded. The algorithms we propose have a high compression rate (5.0 bits/point for the naive distribution and 7.7 bits/point for the layer distribution). Additionally we present the hardware implementation for the decompression of both algorithms. Both algorithms are implemented in a single hardware unit providing a control to select between them.