In this paper decomposition techniques are applied to derivative operators, used for image edge detection. It is shown that the application of decomposition techniques to common edge detectors can result in substantial savings in computing time. For a 25x25 Laplacian of Gaussian, mask, an improvement of six times less arithmetic operations is achieved when decomposition techniques are applied.We also show that these techniques are advantageous for hardware realization of the filters. The memory required to a 2-D (nxn)-th order FIR filter direct realization with distributed arithmetic is O(2(n+1) ) while the worst case for the decomposed filter is O(n x 2n).
