A Radiometric Noise Model for Estimating Geometrical Parameters of 3-D Bodies from Multispectral Images

Camera images can be used to measure the geometry of man-made objects. An iterative weighted least-squares estimator with knowledge of imaging and reflection models retrieves the geometrical parameters of objects in a 3-D scene from 2-D image projections. We investigate the use of multispectral imagery which allows us to separate diffuse and specular reflection before estimating geometry. We built a multispectral camera measurement system that has been used to capture real images of a cylindrical body. These have been processed to analyse the propagation of radiometric noise from reflection via imaging into reflection component separation. The purpose of this research is the development of a radiometric noise model for use in our geometry estimator.