Return to search

Comparison Between Two DifferentScreen Space Ambient OcclusionTechniques

Context. In this project a comparison between two screen space ambientocclusion techniques are presented. The techniques are Scalable AO (SAO)and Multiresolution SSAO (MSSAO) since they both are techniques thatuse mipmaps to accelerate their calculations. Objectives. The aim is to see how big the difference is between the resultsof these two techniques and a golden reference that is an object space raytraced texture that is created with mental ray in Maya and how long timethe computation takes. Methods. The comparisons between the AO textures that these techniquesproduce and the golden references are performed using Structural SimilarityIndex (SSIM) and Perceptual Image Difference (PDIFF). Results. On the lowest resolution, both techniques execute in about thesame time on average, except that SAO with the shortest distance is faster.The only effect caused by the shorter distance, in this case, is that moresamples are taken in higher resolution mipmap levels than when longerdistances are used. The MSSAO achieved a better SSIM value meaningthat MSSAO is more similar to the golden reference than SAO. As theresolution increases the SSIM value between both techniques become moresimilar with SAO getting a better value and MSSAO getting slightly worse,while the execution time for MSSAO has larger increases than SAO. Conclusions. It is concluded that MSSAO is better than SAO in lowerresolution while SAO is better in larger resolution. I would recommendthat SAO is used for indoor scenes where there are not many small geometryparts close to each other that should occlude each other. MSSAO shouldbe used for outdoor scenes with a lot of vegetation which has many smallgeometry parts close to each other that should occlude. At higher resolution,MSSAO takes longer computational time as compared with SAO, while atlower resolution the computational time is similar.

Identiferoai:union.ndltd.org:UPSALLA1/oai:DiVA.org:bth-14867
Date January 2017
CreatorsTörn, Johan
PublisherBlekinge Tekniska Högskola, Institutionen för kreativa teknologier
Source SetsDiVA Archive at Upsalla University
LanguageEnglish
Detected LanguageEnglish
TypeStudent thesis, info:eu-repo/semantics/bachelorThesis, text
Formatapplication/pdf
Rightsinfo:eu-repo/semantics/openAccess

Page generated in 0.0018 seconds