Abstract

v This paper presents a stereo matching approach for ea novel multi-perspective panoramic stereo vision system, mmaking use of asynchronous and non-simultaneous stereo jimaging towards real-time 3D 360? vision. The method is designed for events representing the scenes visual contrast tas a sparse visual code allowing the stereo reconstruction of hhigh resolution panoramic views. We propose a novel cost lmeasure for the stereo matching, which makes use of a simpilarity measure based on event distributions. Thus, the rolbustness to variations in event occurrences was increased. iAn evaluation of the proposed stereo method is presented tusing distance estimation of panoramic stereo views and tground truth data. Furthermore, our approach is compared tto standard stereo methods applied on event-data. Results eshow that we obtain 3D reconstructions of 1024 × 3600 tround views and outperform depth reconstruction accuracy tof state-of-the-art methods on event data. b t t