Investigating Image Fusion Techniques on CHRIS/Proba Space Borne Hyperspectral Data for Material Identification

Abstract

Image fusion can be defined as the process of combining information from two images with unique characteristics to obtain a resultant image which has an assemblage of qualities of both the inputs. Of late, many studies were carried out in fusing multispectral data with panchromatic data, but fusion of hyperspectral data with multispectral data is a topic of interest till date. In the present study, image fusion technique was used to obtain a spatially and spectrally rich hyperspectral (HS) image which can aid in improved material identification. Two conventional pixel basedfusion techniques namely – Gram Schmidt (GS) fusion and High Pass Filtering (HPF) technique were used for fusing the HS CHRIS image with the multispectral (MS) LISS IV image. An overall spectral, spatial/visual and quantitative analysis was carried out to examine the quality offused images. Quantitative analysis of the individual bands was made using SNR and entropy measures. Classification of the fused images helped in identifying five different concrete materials and four different vegetation types within the study area. An overall classification accuracy of 88.33% was obtained using GS fused image, 79.25% with HPF fused image and 73.66% with the CHRIS data.