Investigating the performance of Correspondence Algorithms in Vision based Driver-assistance in Indoor Environment

TL;DR

This study experimentally compares fourteen stereo matching algorithms under various indoor lighting conditions to evaluate their effectiveness for vision-based driver-assistance systems.

Contribution

It provides a comprehensive evaluation of multiple algorithms' performance in indoor environments, highlighting their strengths and weaknesses under different illumination conditions.

Findings

Global algorithms outperform local ones in certain lighting conditions

Algorithm accuracy varies significantly with illumination changes

Some algorithms maintain consistent performance across different indoor lighting scenarios

Abstract

This paper presents the experimental comparison of fourteen stereo matching algorithms in variant illumination conditions. Different adaptations of global and local stereo matching techniques are chosen for evaluation The variant strength and weakness of the chosen correspondence algorithms are explored by employing the methodology of the prediction error strategy. The algorithms are gauged on the basis of their performance on real world data set taken in various indoor lighting conditions and at different times of the day