Sharpening the Sentinel-2A Infrared Bands via Image Residual Optimization

Abstract

Due to the infrared bands, i.e., short-wave infrared (SWIR) and narrow near-infrared (narrow NIR) bands, obtained from the Sentinel-2 multispectral instrument has a lower resolution than the visible-near infrared (VNIR) bands. Therefore, the capability of applications that employ infrared information is often limited by its spatial resolution. To overcome this problem, we propose a method for sharpening these infrared bands. This method is based on the multiresolution analysis (MRA) method, which focuses on the image residual in the spatial domain. The sharpened image can be obtained from the combination of the low-frequency (LF) and high-frequency (HF) components. The residual image is performed as the HF components, which can be estimated from the difference between the original image and its smoothing (blurring) version. The bicubic interpolation was applied to generate the LF component. The least-square error of the image residual was employed to determine the optimal injection weight, which can be expressed in the analytic form. The Sentinel-2A images of Central Thailand were used to test our proposed method and then compared to conventional sharpening methods, such as Gram-Schmitt, Intensity-Hue-Saturation (HIS), and Brovey methods. The results show that our proposed method can give quality metrics higher than others. Moreover, our composited infrared band image also gives color similar to the native composite image.