Automated system for coarse-to-fine pyramidal area correlation stereo matching

摘要

An automatic, general purpose system for the production of dense digital disparity models (DDM) is described. The system applies an area-based correlation stereo matcher via an image resolution pyramid. The theory of this coarse-to-fine technique is described and possible improvements to it are outlined. The results of applying the system to different types of imagery including SPOT-1 stereo imagery of Western Cyprus, structured light stereo imagery of faces and knees and MRI scanner imagery of the human brain are presented.