Difference of Gaussian (DoG).

The Laplacian of Gaussian of Eq. (5.8.7) can be closely approximated by the difference of two Gaussian having property chosen scales. The difference of Gaussian (DoG) filter is

where

and are evaluated by using Eq. (5.8.6). However, the LoG is usually preferred because it is theoretically optimal and its separability allows for efficient computation . For the same accuracy of results, the DoG requires a slightly large filter size.

The technique of unsharp masking, used in photography, is basically a difference of Gaussians operation done with light and negatives. Unsharp masking as discussed earlier involves making a somewhat blurry exposure of an original nagative onto a new piece of film. When the film is developed , it contains a blurred and inverted-brightness version of original negative. Finally, a print is made from these two negatives sandwiched together, producing a sharpened image with edges showing increased contrast.

Nature uses the difference of Gaussians as a basis for the architecture of the retina's visual receptive field. The retina actually implements DoG bandpass filters at several spatial frequencies.