Beyond simulation: assessing 2D IQA metrics in real-world 3D displays and their role in calibration optimization

Objective image quality assessment metrics (IQAs) are essential for evaluating 2D/3D displays, yet they are typically validated on simulated images, overlooking real-world factors such as crosstalk and calibration errors in real-world 3D displays. We systematically tested six leading 2D IQA metrics —PSNR, SSIM, CW-SSIM, GMSD, FSIM, and MDSI— under intentional calibration deviations using a real lenticular autostereoscopic display. Results show that FSIM and MDSI, which employ perceptual features such as phase congruence, excel in robustness and consistency with human perception. Accordingly, this study further proposes an IQA metric based automatic calibration optimization method, validated experimentally. This comprehensive work contributes by: (1) validating IQA metrics in a real stereoscopic environment, (2) identifying FSIM/MDSI for 3D calibration optimization, (3) bridging the simulation-to-reality gap for future IQA metric frameworks, and (4) shifting the paradigm of 3D display calibration from “ physical proofing ” to “perception-driven calibration.”