Image Quality Assessment

Image Quality Assessment asks how good, degraded, or useful an image appears, often when no pristine reference is available, and matters for choosing, improving, or safely using images. Methods range from full-reference metrics such as PSNR and SSIM to no-reference neural predictors, human preference models, and vision-language systems that explain or localize defects. Work increasingly targets specific settings—including generated and restored images, compression, medical scans, and scientific imagery—and uses subjective benchmarks, cross-dataset tests, and downstream task performance. A central tension is that pixel fidelity, human perception, semantic correctness, and task utility can disagree, while models may lose reliability across domains, distortions, or evaluation protocols.