Fusion of High Spatial and High Spectral Data for Quality Enhancement of Remotely Sensed Images

This paper attempts to integrate satellite imagery such as Cartosat-1 with high spatial resolution and IRS P6 LISS IV with high spectral resolution using digital image fusion algorithms. This integration and mixing of hispectral and higher spatial data with complementary spectral and spatial characteristics, has proved to be promising to obtain images with high spatial and spectral resolution simultaneously. This study compares the classification accuracy of original LISS IV imagery and the improvement in classification accuracy after fusion of images. The objective of classification is to identify mangrove species in the Sunderban Biosphere Reserve of West Bengal. The study integrates both imagery using Gram-Schmidt (GS), Principal Component (PC), and Colour Normalised Transform (CNT) algorithms and compares their performance in assessing the classification accuracy with Spectral Angle Mapper in achieving the desired objective. It has been found that the classification results generated after GS fusion gives maximum accuracy (76.92%) followed by PC with 61.53% accuracy and CNT with 53.84% accuracy. It may hence be concluded that the overall accuracy has improved after image integration when compared with the original LISS IV imagery (evaluated as 42.857%). The dominant mangrove species that have been identified in the study area are Excoecaria Agallocha, Avicennia Officinalis, Aegialitis, Avicennia Marina and Ceriops Decandra.