Ground Truth Evaluation for Event-Based Silicon Retina Stereo Data

Abstract

In this paper we present a new approach for the evaluation of event-based Silicon Retina stereo matching results. The evaluation of stereo matching algorithm results is a necessary task for the development, comparison, and improvement of depth generating camera systems. In contrast to conventional frame-based cameras, the silicon retina sensors delivers asynchronous events instead of synchronous intensity or color images. The polarity of the events represents either an increase (on-event) or a decrease (off-event) of the brightness of the projected scene point. This is the reason why existing ground truth data and evaluation platforms are not suitable for testing silicon retina stereo camera systems. For the analysis of the introduced novel evaluation method, we use an area-based (sum of absolute difference) algorithm for the event-driven sensor system. A conventional video camera stereo vision system is used to produce reference data. The results show that the presented method offers new opportunities for the evaluation of stereo matching results computed from silicon retina stereo data.