Perceptual Quality Assessment of Pan-sharpened Images

Pan-sharpening (PS) is an approach to fuse the spatial details of a high-resolution panchromatic (PAN) image and the spectral information of a low- resolution multispectral (MS) image. PS is a preliminary step for enhancing images for remote sensing tasks, such as change detection, object recognition, visual image analysis, and scene interpretation. Given the need for selecting pan-sharpening techniques that provide better spatial and spectral quality of pan-sharpened images, it is highly desirable to be able to automatically and accurately predict pan-sharpened image quality, as would be perceived and reported by human beings and evaluating at the same time spectral distortions as color changes in the PS image. In this research we propose a new image quality assessment (IQA) measure that uses the statistics of natural images, commonly referred to as natural scene statistics (NSS) to extract statistical regularities from PS images. NSS are measurably modified by the presence of distortions, we take advantage of this behavior to characterize some relevant distortions presented in PS images. We analyze six PS methods in the presence of two common distortions, blur, and white noise, on PAN images. Furthermore, we conducted a human study on the subjective quality of pristine and degraded PS images and created a completely blind fused image quality analyzer. In this test, 33 subjects evaluated 420 images in five sessions. In addition, we propose an opinion aware fused image quality analyzer, whose relative predictions with respect to other models match better to human perceptual evaluations than state-of-the-art reduced and full resolution quality metrics. An implementation of the results of subjective study and the proposed fused image quality measures can be found at https://github.com/oscaragudelom/Pansharpening-IQA.