COMPARISON OF LUMINANCE BASED METRICS IN DIFFERENT LIGHTING CONDITIONS

In this study, we evaluate established and newly developed metrics for predicting glare using data from three different research studies. The evaluation covers two different targets: 1. How well the user’s perception of glare magnitude correlates to the prediction of the glare metrics? 2. How well do the glare metrics describe the subjects’ disturbance by glare? We applied Spearman correlations, logistic regressions and an accuracy evaluation, based on an ROC- analysis. The results show that five of the twelve investigated metrics are failing at least one of the statistical tests. The other seven metrics CGI, modified DGI, DGP, Ev, average Luminance of the image Lavg, UGP and UGR are passing all statistical tests. DGP, CGI, DGI_mod and UGP have largest AUC and might be slightly more robust. The accuracy of the predictions of afore mentioned seven metrics for the disturbance by glare lies in the range of 75-83% and does not confirm findings from other studies stating a poor performance of existing glare metrics.


Published in:
Proceedings of the CIE 2017, Midterm Meeting, Jeju Island, Korea
Presented at:
CIE 2017, Midterm Meeting, Jeju Island, Korea, October 20-28, 2017
Year:
2017
Publisher:
CIE, Vienna
Keywords:
Laboratories:




 Record created 2017-11-08, last modified 2018-09-13

Postprint:
Download fulltext
PDF

Rate this document:

Rate this document:
1
2
3
 
(Not yet reviewed)