Extraction of Relief Information on Textured and Coloured Rough Surfaces for the Optimisation of a Propagation Model for Wireless Transmission

• Venue: CRV '04 Proceedings of the 1st Canadian Conference on Computer and Robot Vision
• Year: 2004

This study is carried out within the framework of thedevelopment of simulation systems of high frequencyradio wave propagation (60 GHz) for wireless local areanetworks in an indoor environment. Thus thecorresponding wavelength is in millimeters. At thisfrequency, in order to optimise models of radioelectricwave propagation, it is important to have informationrelated to the 3D roughness of the main reflectivesurfaces encountered during the transmission. But theestimation of the relief of 3D textured surface isgenerally made on grey level images. This supposes thatvariations in grey levels are representative of localvariations in the relief. This assumption is justified in thecase of uniformly coloured surfaces, but is no longervalid when these surfaces present variations of colour oraspect. The corresponding image will then presentvariations in grey levels which can be related to colourvariations or relief variations or both. It becomesdifficult in this case to evaluate relief based on imageanalysis. Before any study of roughness, it is thereforenecessary to devise a method for separating theinformation linked to colour variation from theinformation linked to relief variation. In this paper, wepropose to carry out this separation through aphotometric stereovision system. The method we havedeveloped is based first on the acquisition of threeimages of the studied surface, obtained under differentlight conditions, and second on the photometric model ofthe surface. So we have established the relationshipbetween the local relief of the surface, its colour aspectand the corresponding grey level image. Then, for thestudied surface, we have extracted an imagerepresentative only of its colour aspect and an imagerepresentative only of its local relief. Finally, we haveextended the proposed method to the case of colouredimages.