Interesting approach! I'm curious about your sample image comparison. You mentioned in response to another comment that it was taken by a drone and the image on the right was downsampled by a factor of 3 to mimic the resolution of currently available satellite images. However, the image on the left has artifacting that looks reminiscent of deep learning-based methods for deblurring, denoising, or super-sampling. Is the image on the left actually a raw drone image? How much of the stated 9x improvement (actually 3x) is hardware/sensor-based vs software-based?