Adding functions for calculating covariance from image convolutions. This enables us to have the correct factor to infer the pixel-to-pixel variance in an image following multiple convolutions.