Norman Koren’s concise guide to camera image quality testing for machine vision, centered on information-theory metrics — Information Capacity (C), C4 measured from 4:1 slanted edges, SNRi and Object Detection Error Probability, and the new Information-based Dynamic Range (InfoDR) chart — which predict camera performance better than traditional sharpness or noise metrics.

Christian Taylor and Amelia Limbocker present a method for designing near-infrared (NIR) test charts whose optical densities match a camera’s effective bandpass, extending the ISO 5 visual-density and ISO 14524 OECF models with new spectral weightings to better target NIR machine-vision, automotive, and biomedical applications.

Norman Koren introduces an information-based dynamic range measurement that derives DR and low-light performance from C4 information capacity, measured from 4:1-contrast ISO 12233 slanted edges on a new compact step-square chart, and plots C4 versus exposure as a superior characterization of camera performance across a wide range of illumination.

Sarah Kerr compares centroid, low-pass, and matched-filter edge-localization methods in the ISO 12233 e-SFR algorithm, showing that centroid fitting introduces angular bias under noise while low-pass and matched filtering stay accurate — with direct impact on information-capacity metrics and the emerging ISO/WD 23654 standard.

Norman Koren presents an image sensor noise model — derived from a Photon Transfer Curve (or EMVA 1288 data) and combining dark, photon-shot, and PRNU noise — for use in a full camera/ISP system simulation that predicts both classic metrics (SFR, noise) and new information metrics, most importantly in low light.

Imatest researchers validate printed skin tone test charts against real human subjects spanning the Monk Skin Tone Scale, using CIEDE2000 color-difference analysis to show that wide-gamut face charts enable repeatable, lab-based testing of color accuracy and 3A (auto exposure, white balance, and focus) behavior in face-present scenes.

A University of Galway team, with Valeo and Imatest’s Norman Koren, shows that Shannon Information Capacity (SIC) predicts deep-learning object-detection performance for automated driving far better than traditional metrics such as MTF50 — using a novel simulated test chart of people, cars and cyclists degraded by varying contrast and blur, evaluated across multiple detector architectures.

In 2025, Imatest continued the evolution of our image quality testing capabilities with significant advancements across software, hardware & charts. […]

Norman Koren describes a camera performance simulator that includes an image sensor noise model with parameters derived from either dynamic range test chart measurements or EMVA 1288 results. The simulator includes lens degradations and the effects of Image Signal Proces­sing (ISP). Results can be analyzed for standard image quality metrics such as sharpness and noise, or new metrics, derived from information theory, that provide an improved prediction of machine vision system performance.