DocumentCode
602451
Title
Near surface light source estimation from a single view image
Author
Wu Yuan Xie ; Chung, Chi-kit Ronald
Author_Institution
Mech. & Autom. Eng. Dept., Chinese Univ. of Hong Kong, Hong Kong, China
fYear
2013
fDate
15-17 Jan. 2013
Firstpage
96
Lastpage
101
Abstract
Several techniques have been developed for estimating light source position in indoor or outdoor environment. However, those techniques assume that the light source can be approximated by a point, which cannot be applied safely to, for example, some case of Photometric Stereo reconstruction, when the light source is placed quite close to a small-size target, and hence the size of light source cannot be ignored. In this paper, we present a novel approach for estimating light source from single image of a scene that is illuminated by near surface light source. We propose to employ a shiny sphere and a Lambertion plate as light probe to locate light source position, where albedo variance of the Lambertian plate is used as the basis of the object function. We also illustrate the convexity of this object function and propose an efficient way to search the optimal value, i.e. source position. We test our calibration results on real images by means of Photometric Stereo reconstruction and image rendering, and both testing results show the accuracy of our estimation framework.
Keywords
image reconstruction; photometric light sources; rendering (computer graphics); stereo image processing; Lambertion plate; albedo variance; image rendering; indoor environment; light probe; near surface light source estimation; object function; outdoor environment; photometric stereo reconstruction; real images; shiny sphere; single view image; source position; Estimation; Image reconstruction; Light sources; Lighting; Rendering (computer graphics); Shape; Surface reconstruction;
fLanguage
English
Publisher
ieee
Conference_Titel
Robot Vision (WORV), 2013 IEEE Workshop on
Conference_Location
Clearwater Beach, FL
Print_ISBN
978-1-4673-5646-6
Electronic_ISBN
978-1-4673-5647-3
Type
conf
DOI
10.1109/WORV.2013.6521920
Filename
6521920
Link To Document