imagequality

Image quality describes the perceptual fidelity and usefulness of an image for a given task. It is evaluated relative to a reference or an expected standard and is influenced by factors such as resolution, color accuracy, tonal range, noise, sharpness, and absence of artifacts. The term is used across photography, videography, printing, and display technologies, where higher image quality generally means a more faithful and pleasant representation of the original scene or subject.

Quality assessment can be subjective, using human judgments, or objective, using algorithms. Objective methods are often categorized as full-reference, reduced-reference, or no-reference. Common image quality metrics include PSNR, SSIM, and MS-SSIM, which quantify structural similarity; VIF, DISTS, LPIPS, and Butteraugli, which aim to align with human perception. For video, VMAF is a widely used perceptual metric that extends similar ideas to temporal data.

Subjective testing relies on panels of observers and standardized protocols, producing scores such as mean opinion score (MOS). Standards from ITU-T and other bodies define testing methodologies; ITU-T P.910 covers image quality assessment, while P.913 addresses video quality.

Applications of image quality assessment include optimizing capture pipelines, compression algorithms (JPEG, JPEG2000, HEIF/AVIF, WebP), upscaling and denoising algorithms, display calibration, and quality control in imaging systems. Image quality is also a key consideration in HDR imaging and color management.

Challenges remain in aligning objective metrics with human perception across content types and display conditions. Research continues to improve perceptual models, including deep learning-based metrics, to better predict visual quality under compression, noise, blur, and artifact conditions.