Applicability of White-Balancing Algorithms to Restoring Faded Colour Slides: An Empirical Evaluation

In this paper we investigated the applicability of commonly used white-balancing algorithms to restoring faded photographic colour slides. We have used three sets of synthetic data that simulated colour damage in Kodak Ektachrome slides, as well as three sets of real digitized faded Kodak Ektachrome slides. We have restored all the data sets using nine different algorithms and evaluated restoration results using human participants. In addition, we have conducted an evaluation of the restored synthetic data using a distance metric . We found that while some algorithms provided acceptable restoration of synthetic images, none were found by the human participants to adequately restore real-world data. In addition, we found no correlation between human-based and metric-based evaluation results on 2 out of 3 synthetic data sets. Our results form a strong indication that commonly used white-balancing methods are inadequate for restoring faded colour slides and that simple colour distance metrics do not necessarily correspond to human perception of colour quality.