3D reconstruction of the blood vessels of the brain from a stereoscopic pair of subtraction angiograms

Abstract From physiological considerations about human stereoscopic depth perception we derive an algorithm for automatic image matching of a stereoscopic pair of subtraction angiograms. In a first stage, the blood vessels are seen as strings of micropatterns, characterized by an orientation, width and intensity. Similar micropatterns on the left and the right image of the stereo pair are selected. In a second step, this result is iteratively improved using the consistency property of the blood vessels as a priori knowledge.