Multiplierless coprocessor for difference of Gaussian (DoG) calculation
A hardware architecture is applied to the calculation of a Difference-of-Gaussian filter, which is typically employed in image processing algorithms. The architecture has a modular structure to easily allow the matching of the desired delay/area ratio as well as a high computational accuracy. A new solution is provided for the implementation of multiply-accumulators which allows a significant reduction of area with respect to the conventional architectures.
1. A multiplier accumulator circuit comprising:
a first lookup table configured to provide Bachet terms in response to an input pixel value;
a plurality of second lookup tables configured to provide intermediate values in response to the Bachet terms; and
a set of full adders configured to sum the inte mediate values from the second lookup tables and to provide an output value representative of the input pixel value multiplied by a coefficient.
2. A multiplier accumulator circuit as defined in claim 1 , wherein the set of full adders is configured to add a further input value to the output value representative of the input pixel value multiplied by the coefficient.
3. A multiplier accumulator circuit as defined in claim 1 , configured for use in a Difference-of-Gaussian calculation, wherein the coefficient is a Gaussian coefficient.
4. A multiplier accumulator circuit as defined in claim 1 , wherein the input pixel value is an eight-bit data value.
5. A multiplier accumulator circuit as defined in claim 1 , wherein the plurality of second lookup tables includes six second lookup tables and wherein the set of full adders includes five full adders.
6. A multiplier accumulator circuit as defined in claim 2 , wherein the plurality of second lookup tables includes six second lookup tables and wherein the set of full adders includes six full adders.
7. A multiply accumulate method comprising:
providing Bachet terms from a first lookup table in response to an input pixel value;
providing intermediate values from a plurality of second lookup tables in response to the Bachet terms; and
summing the intermediate values with a set of full adders to provide an output value representative of the input pixel value multiplied by a coefficient.
8. A multiply accumulate method as defined in claim 7 , further comprising adding a further input value to the output value representative of the input pixel value multiplied by the coefficient.
9. A multiply accumulate method as defined in claim 7 , used in a Difference-of-Gaussian calculation, wherein the coefficient is a Gaussian coefficient.