Video Quality Assessment

Video quality assessment studies how well automated scores reflect human judgments of visual fidelity, motion, aesthetics, and—in some settings—content or task success. Methods range from full-reference and no-reference metrics built on handcrafted or learned features to vision-language models, pairwise preference ranking, and motion- or physics-aware evaluation. Subjective datasets and benchmarks cover user-generated content, codecs, streaming, and AI-generated video, often using mean opinion scores or human comparisons for validation. A central challenge is generalizing across content, distortions, devices, and domains while keeping automated judgments aligned with people; objective metrics still do not reliably replace human testing in every setting.