PAN-Sharpening the Spatial Resolution of Multispectral Image for the Assessment of Near-Shore Bathymetry

Abstract

A numerous earth observing satellites have been launched to the orbit to capture the images of the earth surface. The satellite acquires images of earth surface in panchro- matic (Pan) and multispectral (MS) modes. The Pan image contains high spatial details whereas MS image holds rich spectral information but low spatial resolution. The re- mote sensing applications require both the qualities in a single image. Many pan-sharpening methods were developed to transfer the spatial detail of Pan image to the MS image, to have a single image with both high spatial and rich spec- tral information. One of the primary objectives of the present research is to determine the suitable pan-sharpening technique for improving the spatial resolution and retaining the spectral information of MS image. Therefore, the present study focuses on nine different pan-sharpening methods like principal component analysis (PCA), modified- intensity hue saturation (M-IHS), multiplicative, brovey transform (BT), wavelet prin- cipal component analysis (W-PCA), hyperspherical colour sharpening (HCS), high pass filter (HPF), gram-schmidt (GS) and Fuze Go. These were used for fusing the Pan and MS imageries of Quickbird-2 and Landsat-8. The effectiveness of these pan-sharpening method should not distort the spectral information of an MS image while enhancing the spatial resolution. To evaluate the performance of the above mentioned pan-sharpening methods, both qualitative and quantitative approaches were adopted. In the quantitative approach the spectral indices like correlation coefficient (CC), structural similarity index measure (SSIM), root mean square error (RMSE), signal to noise ratio (SNR), universal quality index (Q) and peak signal noise ratio (PSNR) were used to assess the spectral quality of pan-sharpened image. The spatial indices like correlation coefficient (SCC), Gra- dient and image entropy (E) were used to assess the spatial quality of pan-sharpened image. Further, quality with no reference (QNR) indices were also performed to eval- uate both spectral and spatial quality of pan-sharpened image. The results of qualita- tive and quantitative approaches indicate that the Fuse Go method outperformed other pan-sharpening methods in providing a image with the highest spatial details and rich spectral information. In addition, the effectiveness of improving the spatial resolution of MS image was studied by employing the Fuse Go pan-sharpened image for the extraction of buildings using Quickbird-2 imagery and for estimating the bathymetry of near shore ocean using Landsat-8 imagery. To extract the buildings from the original and Fuze Go pan-sharpened Quickbird- 2 imagery, both automatic and manual approaches were adopted and compared using qualitative and metric analysis. In automatic approach firstly, the vegetation portion were removed from the input image. Secondly, adaptive k-means clustering algorithm were adopted to cluster the pixels into different classes. Finally, the morphological fill and open operator was implemented to extract the buildings. In the manual approach, area of interest (AOI) was created from the input image. Later, the generated AOI was used to subset the interested features from the image. The results of qualitative and metric analysis indicate that the building detection percentage of automatic algorithm for the original and pan-sharpened image are reasonable for such a challenging MS im- age. The results of manual method indicate that the extraction of buildings is achieved with minimum loss of information in comparison with the automatic method. How- ever, improving the spatial resolution of the original MS image, helps to determine the buildings information more precisely in terms of spatially as well as spectrally. The procedure based on ratio-transform algorithm was adopted on original and Fuse Go pan-sharpened imagery of Landsat-8 for estimating the bathymtery of near shore ocean along the coast of Mangaluru, India. The performance of the procedure was evaluated using root mean square error (RMSE) and mean absolute error (MAE). The results of RMSE and MAE indicate that the procedure better estimates the depth up to 5 m and 10 m for the original and improved spatial resolution of Landsat-8 imagery. Therefore, the Fuse Go method can be used for remote sensing applications, which demands both high spatial and spectral information in a single image.