Large Texture Storage using Fractal Image Compression

Texture mapping has traditionally added visual realism to computer graphics images. Modern graphic boards reserve more space for image textures that are mapped in real time on 3D meshes. Outside the graphic accelerators texture images are stored in popular compressed formats that are not designed to meet real-time texture decompression requirements. In the paper we forward an alternative technique for image texture handling. We use fractal image compression scheme to decompress very large textures. The main idea is to use high degree of local self-similarity in texture images representing natural scenes. The method delivers very high compression ratio for complex textures and close to real-time decompression. Local compression schemes allow for Region Of Interest access to the visual data. Properties of the method are compared to those of classical DCT and wavelet compression schemes.