Stereo analysis of archaelogical scenes using monogenic signal representation

Abstract

This paper presents the results of an experimental study regarding the application of recent stereo analysis theories in the frequency domain, particularly the phase congruency and monogenic filtering methods. The initial approach to the stereo matching problem employed feature based correlation methods. However, the requirement for more dense depth-map output led us to the development of disparity map estimation methods, minimizing a matching cost function between image regions or pixels. The cost function consists of a newly proposed similarity measure function, based on the geometrical properties of the monogenic signal. Our goal was to examine the performance of these methods in a stereo matching problem setting, on photos of complicated scenes. Two objects were used for this purpose: (i) a scene from an ancient Greek temple of Acropolis and (ii) the outside scene of the gate of an ancient theatre. Due to the complex structure of the photographed objects, classic techniques used for stereo matching give poor results. On the contrary, the three-dimensional models and disparity map of the scene computed when applying the proposed method, are much more detailed and consistent. © 2010 Springer-Verlag Berlin Heidelberg.