Image Quality Evaluation of Various Pan-Sharpening Techniques Using Landsat-8 Imagery

Abstract

A widely used image processing technique called Pan-sharpening (PS) can be used for fusing the spectral details from a Multispectral (MS) image with a high spectral and low spatial resolution (such as Landsat multi-band image) with the spatial details of High Resolution (HR) Panchromatic (PAN) image (such as Landsat PAN band), for generating high spectral and spatial resolution MS images. Due to the increasing availability of different sensors, PS using single-source or multi-source satellites is gaining more attention among Remote Sensing (RS) researchers. Pan-sharpening techniques mostly vary in how well they integrate spectral and spatial details into the fused image, and hence their efficiency also varies. This study evaluated the efficiency of eight different PS techniques using Landsat-8 imagery which will help in choosing the best PS technique based on their accuracies for particular applications. These techniques include the High Pass filter Additive (HPFA), Hue-Saturation-Value (HSV) transform, Gram-Schmidt (GS), Intensity Hue Saturation (IHS), Smoothing Filter based Intensity Modulation (SFIM), Principal Component Analysis (PCA), Simple Mean (SM), and Brovey Transform (BT). The image quality of these PS techniques is assessed using ten different metrics such as Erreur Relative Globale Adimensionnelle de Synthèse (ERGAS), Root Mean Squared Error (RMSE), bias, Difference in Variance (DIV), Pearson’s correlation coefficient (PCC), Universal Image Quality Index (Q), Change in Mean Contrast (CMC), Relative Average Spectral Error (RASE), Change in Mean Luminance (CML), and Mean Squared Error (MSE). The spatial improvement was verified qualitatively and quantitatively. From the whole pan-sharpened (fused) images, urban characteristics including streets, buildings, water bodies, and the stadium could be easily recognized. Qualitative analysis revealed that the GS and SFIM techniques generated the best pan-sharpened images while poor results were obtained by the PCA technique.