@MASTERSTHESIS\{IMM2014-06830, author = "V. Vilcans", title = "GPU-Based Global Illumination - Vertex Connection and Merging using OptiX", year = "2014", school = "Technical University of Denmark, Department of Applied Mathematics and Computer Science", address = "Richard Petersens Plads, Building 324, {DK-}2800 Kgs. Lyngby, Denmark, compute@compute.dtu.dk", type = "", note = "{DTU} supervisor: Jeppe Revall Frisvad, jerf@dtu.dk, {DTU} Compute", url = "http://www.compute.dtu.dk/English.aspx", abstract = "Physically based rendering algorithms are capable of producing high quality images of virtual environments. The rendering time, however, usually is long due to associated computation costs. The {GPU} processing power is increasing in a relatively faster pace than CPUs and they are specially designed to handle parallelizable workloads. This makes them attractive for accelerating physically based algorithms. In recent years multiple tools such as OpenCL, Nvidia's {CUDA} and OptiX for harnessing {GPU} power have appeared. The goal of the thesis is to provide an overview of the novel Vertex Connection and Merging (VCM) [GSK11] algorithm which combines benefits of Bidirectional Path Tracing and Photon Mapping, and to describe an implementation using Nvidia's OptiX ray-tracing framework [PBD+10]. The taken approach was to extend the OppositeRenderer, an open source OptiX based renderer which contains Progressive Photon Mapping (PPM) implementation. We hoped that would help to speed up the implementation of Vertex Merging through partial reuse of data structures used in {PPM} and to provide richer feature set in the end. Due to experienced development issues and delays, Vertex Merging was not implemented, however, Vertex Connections including the common parts for both path sampling methods were implemented. Also material handling capabilities were improved to allow flexibility of combining multiple reflection models for single material. The extended version of Opposite Renderer is provided to the public same as the original was." }