HDR-VDP-2, LPIPS, DISTS, and SSIM for Evaluating Computer-Generated Renderings: A Systematic Review of Use and Perceptual Validation

Abstract

This systematic review examined the use and perceptual validation of four full-reference image quality metrics—HDR-VDP-2, LPIPS, DISTS, and SSIM—in computer-generated and rendered imagery. Among 283 included studies, SSIM was reported in 228 (80.6%) and LPIPS in 147 (51.9%), whereas HDR-VDP-2 and DISTS appeared in four each (1.4%). Only 23 studies (8.1%) included relevant human or subjective evaluation, and 13 (4.6%) directly tested metric–human agreement. Neural rendering, novel-view synthesis, and 3D reconstruction accounted for 175 studies (61.8%). Metric–human agreement varied with content, artifact type, alignment, viewing conditions, and perceptual task, while cue-specific evidence was uneven. The review therefore identifies a clear gap between metric adoption and perceptual validation. Metric selection should be matched to the rendering attribute, reference image, viewing conditions, and perceptual question.