Overview
ISO/IEC 29170-3:2026, titled Information technology - JPEG AIC Assessment of Image Coding - Part 3: Subjective Quality Assessment of High-Fidelity Images, is an international standard from ISO and IEC. This document establishes a robust framework for subjective assessment of image quality in high-fidelity ranges, particularly targeting images that range from good quality up to mathematically lossless compression. The focus is specifically on evaluating distortions arising from image coding processes, such as those introduced by lossy compression algorithms, rather than from image capture, sensor, or rendering artefacts.
This standard provides practical methodologies for developers, testers, and researchers to carry out psychophysical assessments using human observers, ensuring precise and reproducible quality evaluation for advanced image compression codecs.
Key Topics
- Subjective Image Quality Assessment: Details experimental procedures focusing on human perception of image fidelity (how closely a compressed image matches the original).
- Fidelity vs. Appeal: The methodology emphasizes assessment of fidelity-truthfulness to the source image, rather than aesthetic appeal.
- Triplet Comparison Approach: Introduces a method where two distorted images and their original source (“pivot”) are presented for comparative evaluation by observers.
- Boosted and Plain Triplet Comparisons:
- Boosted Techniques: Use artefact amplification, zooming, and flickering to enhance visibility of distortions, enabling fine-grained differentiation, especially near visually lossless quality.
- Plain Techniques: Rely on unmodified image presentations for baseline comparison.
- Observer Selection and Viewing Conditions: Specifies criteria for selecting observers with normal visual acuity and color vision, and outlines both controlled laboratory and crowdsourced test environments.
- Dataset Preparation: Recommends selection of high-quality, diverse, representative source images appropriate to the codecs and targeted applications.
- Statistical Data Analysis: Employs psychometric scaling, notably Thurstone’s Case V model and Maximum Likelihood Estimation (MLE), to quantify perceptual differences in “just noticeable difference” (JND) units.
- Data Cleansing and Quality Control: Includes robust procedures for filtering out inconsistent or unreliable observer responses.
Applications
The standard’s subjective quality assessment methodology is highly relevant in scenarios such as:
- Development and Optimization of Image Codecs: JPEG or similar image compression algorithms can be objectively compared at high-fidelity levels to ensure minimal perceivable loss.
- Benchmarking and Quality Assurance: Provides a consistent approach to assess the visual performance of different codecs and configurations, supporting procurement, R&D, and product development decisions.
- Research in Visual Perception and Coding: Facilitates academic studies on human visual thresholds for compression artefacts, supporting advancements in perceptual coding.
- Medical and Scientific Imaging: Assists in evaluating compression schemes for applications requiring preservation of visually lossless image quality.
- Standardization and Compliance Testing: Supports regulatory or quality certification processes by enabling repeatable, recognized methods for high-fidelity image assessment.
Related Standards
Organizations implementing ISO/IEC 29170-3:2026 may benefit from familiarizing themselves with these related standards:
- ISO/IEC 29170-1: General framework for JPEG AIC assessment of image coding.
- ISO/IEC 29170-2: Procedures for visually lossless fidelity assessment, including the flicker test used as a methodological basis in Part 3.
- ITU-R BT.500: Referenced for protocols on subjective visual assessment of quality, especially in controlled viewing conditions.
- ISO 8596: Visual acuity testing (relevant to observer selection procedures).
By adhering to ISO/IEC 29170-3:2026, stakeholders in the information technology and imaging sectors can ensure that their subjective image quality evaluations for high-fidelity images are accurate, consistent, and internationally recognized.