Band grouping: Leveraging Spectral Correlation for Improved HSI Processing

The correlations obtained can be used as a technique for grouping bands to leverage information from adjacent pixels. High spectral density in hyperspectral images (HSI)s can cause computational complexity. We demonstrate that grouping band information improves generalizations while requiring less computation and strengthens interpolation learning. The bands adjacent to each other display high spectral similarity that is captured by a statistical metric such as correlation matrix. We investigate the impact of combining related spectral bands as a cohesive strategy for efficient interpolation. This addition of detail and texture cannot be ignored because it helps in better resolved images.