BIQ2021: a large-scale blind image quality assessment database

The assessment of the perceptual quality digital images is becoming increasingly important as a result widespread use multimedia devices. Smartphones and high-speed internet are just two examples technologies that have multiplied amount content available. Thus, obtaining representative dataset, which required for objective training, significant challenge. Blind Image Quality Assessment Database, BIQ2021, presented in this article. By selecting with naturally occurring distortions reliable labeling, dataset addresses challenge no-reference image assessment. consists three sets images: those taken without intention using them assessment, intentionally introduced natural distortions, from an open-source image-sharing platform. It attempted to maintain diverse collection various devices, containing variety different types objects varying degrees foreground background information. To obtain scores, these subjectively scored laboratory environment single stimulus method. database contains information about subjective scoring, human subject statistics, standard deviation each image. dataset's Mean Opinion Scores (MOS) make it useful assessing visual quality. Additionally, proposed used evaluate existing blind approaches, scores analyzed Pearson Spearman's correlation coefficients. MOS freely available benchmarking.