Predicting the Perceptual Quality of Visible and Long-Wave Infrared (LWIR) Images
Abstract: The role of image quality assessment in tasks such as (i) the fusion of long wave infrared (LWIR) and visible images and (ii) face recognition in LWIR images has not been researched extensively from the natural scene statistics (NSS) perspective. For instance, even though there are several well-known measures that quantify the quality of fused images, there has been little work done on analyzing the statistics of fused LWIR and visible images and associated distortions. Furthermore, there is a lack of studies on the influence of degradation of image quality on the performance of automatic face recognition on LWIR images. In this talk, we will present an opinion-aware (OA) fused image quality analyzer, whose quality prediction power exceeds that of other state-of-the-art picture quality models in regards to correlation with human subjective judgments. Moreover, we quantify the impact of common image distortions on infrared face recognition, and present a method for aggregating perceptual quality-aware features to improve recognition rates. We use NSS to detect degradation of infrared images, and to adapt the face recognition algorithm to the quality of the test image. The proposed approach applied to a face identification algorithm based on thermal signatures yielded an improvement of rank one recognition rates between 11% and 17%. These results confirm the relevance of NSS for improving non-reference quality evaluation of used images and to biometric identification systems that use thermal images.