An Enhanced HDR Video Quality Assessment Approach Through Luminance Clipping Detection

This work focused on the limitations of current full-reference video quality assessment models when applied to High Dynamic Range (HDR) content. Most traditional metrics, including fusion-based methods such as FUNQUE+, are pri-marily designed for Standard Dynamic Range (SDR) and High Definition (HD) video. As a result, they do not adequately account for luminance behavior and highlight distortions that are crucial for HDR perception. To address this limitation, an HDR-aware extension of the FUNQUE+ framework was pro-posed, incorporating a luminance clipping feature to capture brightness saturation and overexposed regions in PQ-encoded video. Evaluation on HDR content showed consistent improve-ments in correlation with subjective quality scores compared to the original FUNQUE+ framework. These gains confirm that considering HDR-specific luminance distortions leads to improved perceptual alignment, particularly for sequences affected by strong highlight saturation. In summary, the proposed HDR-extended FUNQUE+ framework provides a more reliable and perceptually mean-ingful approach for assessing modern high-luminance video content.

[1]     Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment: From error visibility to structural similarity,” IEEE Transactions on Image Processing, vol. 13, no. 4, pp. 600–612, 2004. [2]     M. Sibai, D. Temel, and G. AlRegib, “What is HDR? Perceptual impact of luminance and contrast in high dynamic range video,” ACM Trans. Multimedia Comput. Commun. Appl., vol. 15, no. 1s, pp. 1-21, 2019. [3]     Z. Wang, F. Gao, and A. C. Bovik, “Overview of high-dynamic-range image quality assessment,” Frontiers in Neuroscience, vol. 18, 2024. doi:10.3389/fnins.2024.xxxxxx [4]     R. Mantiuk, K. J. Kim, A. G. Rempel, and W. Heidrich, “HDR-VDP-2: A calibrated visual metric for visibility and quality predictions in all luminance conditions,” ACM Trans. Graph., vol. 30, no. 4, Article 40, 2011. [5]     Netflix, “VMAF: The new standard for video quality assessment,” [Online]. Available: https://github.com/Netflix/vmaf [6]     Y. Wang, S. Wang, and S. Ma, “HDR-VQM: An objective quality mea-sure for high dynamic range video,” Signal Process. Image Commun., vol. 75, pp. 52-62, 2019. [7]     J. Alakuijala, et al., “HDRMAX: A framework for subjective and ob-jective quality assessment of HDR video,” IEEE Trans. Image Process., vol. 30, pp. 4409-4424, 2021.15, no. 1s, pp. 1-21, 2019. [8]     D. Ghadiyaram, J. Pan, and A. C. Bovik, “LIVE HDR video qual-ity assessment database,” The Univ. Texas at Austin, 2015. [On-line]. Available: https://live.ece.utexas.edu/research/LIVEHDRVideo QualityDatabase.html [9]     J. Chen, X. Liu, T. Bai, et al., “HDRSDR-VQA: A subjective video quality dataset for HDR and SDR content under realistic viewing conditions,” arXiv preprint arXiv:2312.06411, 2023. [10]    Y. Zhu, et al., “HIDRO-VQA: High dynamic range oracle for video quality assessment,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), 2023.