@MASTERSTHESIS\{IMM2013-06658,
    author       = "J. Udsholt",
    title        = "Real-Time Rendering of Procedurally Generated Volumetric Models",
    year         = "2013",
    school       = "Technical University of Denmark, Department of Applied Mathematics and Computer Science / {DTU} Co",
    address      = "Matematiktorvet, Building 303B, {DK-}2800 Kgs. Lyngby, Denmark, compute@compute.dtu.dk",
    type         = "",
    note         = "{DTU} supervisors: Jakob Andreas B{\ae}rentzen, janba@dtu.dk, {DTU} Compute, and Jeppe Revall Frisvad, rerf@dtu.dk, {DTU} Compute",
    url          = "http://www.compute.dtu.dk/English.aspx",
    abstract     = "This thesis presents a volume based approach to real-time rendering of highly detailed procedurally generated scenes. This approach is based on existing methods for sparse voxel octrees and cache based data management, which aims to solve the problem of maintaining large volume data sets in memory.
Two extensions, designed to provide an efficent framework for procedural content generation in a parallel environment, are suggested: The first is a three step method for subdivision of sparse voxel octree nodes, when the contents of the node is unknown prior to evaluation. The second is a simplification to cache invalidation of unused octree nodes.
A hardware accelerated volume rendering implementation is presented. It uses NVIDIAs {CUDA} platform to perform volume ray casting in real-time on consumer graphics hardware. The implementation is evaluated based on performance, memory requirements and image quality. Performance bottlenecks related to procedural content generation and scalability are identified, and possible solutions are presented. Finally, potential areas for further work are outlined."
}