My diploma thesis at TU Dresden (in cooperation with Crytek GmbH) achieved 50% better compression than GZIP for 3D models, enabling efficient streaming for web-based 3D applications.
As 3D content moved to the web with WebGL, loading times became critical. The research evaluated compression methods and streaming techniques to transmit models and textures efficiently while maintaining quality.
I evaluated multiple compression formats that achieved 2:1 improvement over GZIP binary, implemented progressive mesh streaming for early preview while downloading, and tested Level of Detail (LOD) techniques for adaptive quality. For textures, I implemented progressive JPEG for incremental loading and tested the image geometry technique for efficient compression. The performance analysis showed that decode time becomes the bottleneck on fast connections.
The practical recommendation: for fast connections (>10 Mbps), native texture loading outperforms progressive streaming due to decode overhead. For slower connections, progressive JPEG provides better user experience. I also evaluated client-side caching strategies using IndexedDB and LocalStorage.
This research informed best practices for web-based 3D applications like CAD viewers, product configurators, architectural visualization platforms, and web-based game engines. The compression techniques I evaluated became foundation knowledge for optimizing 3D content delivery in later projects.
The work covered WebGL Loader, OpenCTM, Progressive Meshes (Hoppe et al.), progressive JPEG/PNG streaming, and decode performance benchmarks. Built with WebGL, JavaScript, C++/Emscripten, and various 3D geometry processing libraries.