Multispectral Image Super Resolution with Auto-Encoder Model and Fusion Technique

Obtaining High Resolution(HR) Multispectral Images which are not readily available is one of the more critical objectives in remote sensing applications as these images can be used for various agricultural applications and previously various other methods like pansharpening have been introduced. This paper proposes a novel convolutional auto-encoder for training the multispectral images obtained from Sentinel -2A Satellite and then pass the degraded multispectral image to obtain the reconstructed Multispectral Image which is spectrally enhanced and then fuse the image obtained from reconstruction with the original degraded image to obtain a spatial HR Multispectral Image. This fusion is done using various state of the art methods like Principal Component Analysis(PCA), Discrete Wavelet Transform Level-l(DWT) and Stationary Wavelet Transform Level-l(SWT) and the performance metrics.