derivative of convolution derivative of the gaussian DoG = Laplacian of Gaussian LoG instead of doing 2 convolutions, we can just do 1 to get the same result filter extracts useful info
