Fast Obstacle Detection Using Sparse Edge-Based Disparity Maps

Abstract

This paper presents a fast approach for computing image stixels from a sparse edge-based disparity map. The use of edge-based disparity maps speeds up the computation of the stixels as only a few pixels must be processed compared to approaches which use dense disparity maps. The proposed approach produces as output the stixels in one of the views of the stereo-pair and a segmentation of the edge-points into obstacle. Additionally the proposed approach allows the identification of partially occluded objects by allowing more than one stixel per image column. The proposed approach is fast to compute with no loss on accuracy.