Evaluation of image quality metrics designed for DRI tasks with automotive cameras

Abstract

Nowadays, cameras are widely used to detect potential obstacles for driving assistance. The safety challenges have pushed the automotive industry to develop a set of image quality metrics to measure the intrinsic camera performances and degradations. However more metrics are needed to correctly estimate computer vision algorithms performance, which depends on environmental conditions. In this article we consider several metrics that have been proposed in the literature: CDP, CSNR and FCR. We show a test protocol and promising results for the ability of these metrics to predict the performance of a reference computer vision algo-rithm that was chosen for the study.