Image Quality Testing: Selection of Images for Assessing Test Subject Input

Abstract

Determining image quality is dependent to some degree on human interpretation. Although entirely subjective methods of evaluating image quality may be adequate for consumer applications, they are not acceptable for security and safety applications where operator interpretation may lead to missing a threat or finding threats where they do not exist. Therefore, methods must be developed to ensure that the imagery used in security and safety applications are of sufficient quality to allow the operator to perform his job accurately and efficiently. NIST has developed a method to quantify the capability of imagers to provide images of sufficient quality to allow humans to perform specific perception-based tasks. A one-time humanperception based step is required that results in perception coefficients that are combined with lab-measured objective image quality indicators (IQIs) to calculate image quality. This work uses a d′ evaluation method to examine the performance of test subjects in the human-perception based step, which was identification of a fire hazard in a set of grey-scale infrared images.