Please use this identifier to cite or link to this item: https://www.um.edu.mt/library/oar/handle/123456789/24640
Title: Efficient rendering of soft shadows for multiple dynamic area light sources
Authors: Griscti Soler, Nicholas
Keywords: Computer graphics
Approximation algorithms
Many-body problem -- Computer simulation
Issue Date: 2017
Abstract: Realistic soft shadows, cast by area and volumetric light sources, are highly sought after in computer graphics. However, these are amongst the most computationally expensive to render due to complexities in penumbra calculations. In this final year project, a novel approach based on area light point-sampling is pro- posed which trades off visual quality for rendering performance. The approach is suitable for area lights which emit light uniformly along their surface (most light sources fall under this category). Sampling is performed independently of the area light geometry by making use of depth and normal maps taken around the area light. These textures are then sampled to reconstruct surface positions on the area light geometry. The cost of rendering time is then reduced by clustering reconstructed sample points using a heuristic which takes in consideration the proximity of sampled points to the camera. This is achieved using an extended implementation of the Barnes-Hut algorithm (approximation method for the N-Body problem in astrophysics). Visual quality and performance metrics are used to evaluate the proposed method across a variety of area light sources and scenes. Results show an improvement in performance as the number of samples decreases. This usually comes at the cost of quality, however, sufficiently good images can still be achieved with fewer samples. This trade-off allows for the use of area light sources at interactive rates.
Description: B.SC.(HONS)COMP.SCI.
URI: https://www.um.edu.mt/library/oar//handle/123456789/24640
Appears in Collections:Dissertations - FacICT - 2017
Dissertations - FacICTCS - 2017

Files in This Item:
File Description SizeFormat 
17BCS016.pdf
  Restricted Access
175.72 MBAdobe PDFView/Open Request a copy


Items in OAR@UM are protected by copyright, with all rights reserved, unless otherwise indicated.