期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2008
卷号：XXXVII Part B3a
页码：63-68
出版社：Copernicus Publications
摘要：Global algorithms currently represent the state-of-the-art in dense stereo matching. These methods first set up an energy function. The energy function is then subject to optimization, which is typically achieved via graph-cuts or belief propagation. In this paper, we concentrate on the energy modelling aspect. An experimental study that focuses on the role of colour in stereo energy functions is presented. We evaluate the performance of various forms for using colour and compare it against grey-scale matching. Colour is thereby represented in nine different colour systems. The L 1 and L 2 distances are evaluated for computing the colour differences in the selected systems. We embed the resulting energy functions into two stereo algorithms and test them on 30 ground truth test image pairs. The results of our benchmark show that colour information, in general, leads to a significant performance gain over using intensity only. According to our evaluation results, the selection of the applied colour space is of specific importance in global stereo matching
