Directional statistics-based quality measure for spotlight color images

Abstract

The present paper addresses a new problem related to measuring the quality of color spotlight images. Its primary aim is to raise the limitation problem of available cameras to reproduce the spotlight colors, especially at night. To address this issue, a new reduced reference quality measure is proposed to measure the spotlight color degradation. The idea focuses on transforming the color information into another space where it is defined as an orientation represented on a unit sphere. Then, the directional statistics-based von Mises–Fisher probability density function is used as a deviation measure. To validate the proposed model, a new collection of widely used spotlight color images is constructed. The collection contains a hundred of spotlight colors captured by different cameras in Sherbrooke city as well as available images on the Web. Obtained results are promising.