On Round-Off Errors and Gaussian Blur in Superresolution and in Image Registration