Fusion of hyperspectral and multispectral images based on principal component analysis and guided bilateral filtering

Spectral and spatial resolutions play a vital role in remote sensing applications. However, due to the limitations of imaging sensors, hyperspectral image (HSI) with good spectral features often suffers from poor spatial information. To address this problem, HSIs are to be fused with their multispectral image (MSI) versions. Image fusion is the combination of multiple images of same scenes to intensify salient features in the fused image. It is widely used in agriculture, medical, remote sensing areas. In our proposed method, a unique edge-preserving HSI-MSI fusion is developed using principal component analysis (PCA) and guided bilateral filter (GBF). PCA eliminates the correlated variables and increases the variance. The HSI is spatially improved by replacing with the highest variance principal component with its MSI. In addition, the cascaded GBFs present restore the edge details in the fused image. Using three reference and four non reference public datasets, the effectiveness of our method is demonstrated over the existing methods. We have reported 36.98 dB peak signal-to-noise ratio and 0.764 universal image quality index, which are averaged over three HSI-MSI datasets.