Repository hosted by TU Delft Library

Home · Contact · About · Disclaimer ·

A comparative study of fast dense stereo vision algorithms

Publication files not online:

Author: Sunyoto, H. · Mark, W. van der · Gavrila, D.M.
Source:2004 IEEE Intelligent Vehicles Symposium, 14-17 June 2004, Parma, Conference code: 63502, 319-324
IEEE Intelligent Vehicles Symposium, Proceedings
Identifier: 238010
Keywords: Informatics · Algorithms · Data acquisition · Error analysis · Intelligent vehicle highway systems · Optimization · Program processors · Surveying · Traffic control · Computational costs · Single instruction multiple data (SIMD) · Stereo algorithms · Synthetic datas · Stereo vision


With recent hardware advances, real-time dense stereo vision becomes increasingly feasible for general-purpose processors. This has important benefits for the intelligent vehicles domain, alleviating object segmentation problems when sensing complex, cluttered traffic scenes. In this paper, we present a framework ofreal-time dense stereo vision algorithms all based on a SIMD architecture. We distinguish different methodical components and examine their performance-speed trade-off. We furthermore compare the resulting algorithmic variations with an existing public source dynamic programming implementation from OpenCV and with the stereo methods discussed in Sharstein and Szeliski's survey. Unlike the previous, we evaluate all stereo vision algorithms using realistically looking simulated data as well as real data, from complex urban traffic scenes.