Texture analysis and repacking for improved storage efficiency

Abstract

Textures are widely used in modern computer graphics. Their size, however, is often a limiting factor. Considering the widespread adaptation of mobile virtual and augmented reality applications, efficient storage of textures has become an important factor. We present an approach to analyse textures of a given mesh and compute a new set of textures with the goal of improving storage efficiency and reducing memory requirements. During this process the texture coordinates of the mesh are updated as required. Textures are analysed based on the UV-coordinates of one or more meshes and deconstructed into per-triangle textures. These are further analysed to detect single coloured as well as identical per-triangle textures. Our approach aims to remove these redundancies in order to reduce the amount of memory required to store the texture data. After this analysis, the per-triangle textures are compiled into a new set of texture images of user defined size. Our algorithm aims to pack texture data as tightly as possible in order to reduce the memory requirements.