By calculating the 2-D cumulative sumation, I’ve pre-calculated much of what will be needed to determine the convolutions. I’ll will have reduced every sum over a kernel to an operation over four numbers.