Evaluating Color Difference Models for Better Color Reproduction
Author Information
Author(s): Basova Olga, Gladilin Sergey, Kokhan Vladislav, Kharkevich Mikhalina, Sarycheva Anastasia, Konovalenko Ivan, Chobanu Mikhail, Nikolaev Ilya
Primary Institution: Federal Research Center “Computer Science and Control” of the Russian Academy of Sciences
Hypothesis
Can color difference models be validated effectively in wide color gamut and high dynamic range contexts?
Conclusion
The study found that while the CAM16-UCS model with power correction performed well, it still requires further refinement to improve accuracy.
Supporting Evidence
- The study developed a new dataset covering a wide range of color differences.
- Benchmarking revealed that CAM16-UCS with power correction is the most versatile model.
- Significant errors were still present even in the best-performing model.
Takeaway
This study looked at how well different models can tell the difference between colors, especially in bright and colorful images, and found that improvements are still needed.
Methodology
A new dataset was created to test color difference models, and various models were benchmarked using psychophysical data.
Potential Biases
Some biases may have been introduced by the experimental setup and observer variability.
Limitations
The best model still produced significant errors and was not trained on HDR or WCG data.
Participant Demographics
22 observers with normal or corrected vision participated in the study.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website