Tech Reports Home Contact
Technical Report Series
Actions
Submit Report
Browse Reports
Information
Information for Authors
Contact Details

Technical Report HW-MACS-TR-0020


TitleHow do joint image statistics change with illumination?
AuthorsOndrej Drbohlav, Mike Chantler
Date2004-08-11
AbstractThe dependence of statistical properties of an image as a function of illumination direction is an exciting topic which was so far investigated experimentally, and also addressed theoretically for a special case of low-order moments of image features. In this paper we observe the principal relationship between the joint (co-occurrence) distribution of surface properties and the corresponding joint distribution of local image features. We focus on two kinds of statistics computed from local image neighbourhoods: (a) the joint distribution of pixel intensities, and (b) the joint distribution of binary patterns obtained by taking the signs of intensity differences between a selected reference pixel and all other pixels. We work with a non-parametric histogram representation of the probability distribution functions, and show how the frequencies in a histogram bin depend on illumination. Finally, we focus on approximating the illumination dependence of image statistics by a harmonic series. Experimental results obtained using a real surface are presented.
GroupVision
Notes
Download

 

Email Technical Report's Administrator
|MACS Home| Top of the Page

Department of Computer Science, Heriot-Watt University, Riccarton, Edinburgh, EH14 4AS, +44 (0) 131 4514152

Last Updated: 02 September 2003 Copyright Heriot-Watt University, Disclaimer