Natural Scene Statistic

Natural scene statistics (NSS) describe the statistical regularities found in naturally occurring images, informing image quality assessment and computer vision tasks. Current research focuses on adapting NSS models to handle the increasing volume of user-generated content and screen-captured images, which deviate from traditional NSS assumptions, often employing deep learning techniques to learn these new statistical regularities or incorporating saliency maps to improve perceptual quality assessments. This work is crucial for improving the accuracy and robustness of image quality metrics and computer vision algorithms, particularly in applications involving diverse and potentially noisy image sources.