Picture Quality of Security Relevant Images
Ensuring the security of public transportation portals, company facilities, and government installations has become a topic of increased concern over the past 15 years. Thus, it becomes important to develop new and diverse imaging modes capable of conducting screening and inspection processes under increasingly difficult and busy conditions, as well as to test and verify the efficacy of the new and developing imaging devices that use these diverse and various modes.
WNCG Professor Alan C. Bovik and his students are working on building automatic prediction models that can assess the perceptual quality of diverse modes of images, such as X-Ray, microwave, passive millimeter-wave, Infrared, and other modalities, with an aim towards understanding the imaging efficacy of the devices that deliver the images.
To achieve these goals, the research team is expanding single pixels and multivariate models of natural images (NSS) models beyond visible light images. They showed that these tools predict human performance for assessing distortion and quality in long wave infrared images very accurately. The next step would be to expand these methods to span the above mentioned modalities and predict the quality of security cameras over different capturing conditions to accomplish different kinds of security tasks.
To this end, the team is planning a series of picture quality assessment studies that will measure the task performance of law enforcement personnel seeking specific signatures of objects in images, such as weapons, explosive devices and so on, as a function of picture quality.
The study will analyze important aspects of image quality for task performance and determine the relationship between conventional image quality metrics (IQM) and the library of “intelligent” NSS perceptual quality-aware models.
This work is supported by The National Institute of Standards and Technology (NIST)