Spatio-Spectral Image Fusion Using Local Embeddings

Image fusion is extensively used in remote sensing to combine multiple images into a more informative single image, which is more suitable for human and machine perception. The purpose of image fusion algorithms is to achieve a single image with high spatial resolution and high spectral resolution. The major challenges faced by existing algorithms are of output image quality in terms of colour distortion and blurring. This paper models the spatio-spectral image fusion as a problem of non-linear dimensionality reduction. The aim is to define each point of fused image as a linear combination of its neighbours in multi spectral image and the corresponding pixel values from down scaled panchromatic image. The assumption here is that the spectral behaviour of fused image is similar to that of multispectral image. The performance of the model is compared with existing state of the art methods for same-sensor dataset.