Abstract: This disclosure relates to image processing and in particular to super-resolution of mosaic images. An image processor receives a low-resolution mosaic input image and applies a trained neural network to the low-resolution mosaic input image to generate output data of the neural network. Then, the processor creates a high-resolution mosaic output image based on the output data of the neural network. Each of the low-resolution mosaic image and the high-resolution mosaic image is represented by two-dimensional image data comprising, for each pixel, intensity values of different spectral bands arranged adjacent to each other in two dimensions such that each intensity value has a unique location in the two-dimensional image data.

Abstract: This disclosure relates to image processing and in particular to super-resolution of mosaic images. An image processor receives a low-resolution mosaic input image and applies a trained neural network to the low-resolution mosaic input image to generate output data of the neural network. Then, the processor creates a high-resolution mosaic output image based on the output data of the neural network. Each of the low-resolution mosaic image and the high-resolution mosaic image is represented by two-dimensional image data comprising, for each pixel, intensity values of different spectral bands arranged adjacent to each other in two dimensions such that each intensity value has a unique location in the two-dimensional image data.