Colour segmentation under illumination fluctuations

Abstract

In low level image analysis of a 3D scene, illumination effects like shadows, shading, inter reflections and highlights can severely degrade system performance. These effects create erroneous edges and regions in traditional Intensity based methods and complicate higher level recognition process. In this thesis, colour segmentation is used to minimise these illumination fluctuations without a priori knowledge and models of complex geometrical shapes.