An Efficient FPGA Implementation of Approximate Multiply Accumulate Unit for Image and Video Processing Applications in Healthcare Sector
P.L. Lahari, Rahul Gowtham Poola, Siva Sankar Yellampalli · 2024
This chapter discusses the effective approximate multiply accumulate (MAC) unit for image and video processing applications. The MAC unit consists of the multiplier, adder, and accumulator which use the various multipliers such as array, Wallace, Booth, and modified Booth. A modified booth multiplier has the benefit of accepting inputs of both signed and unsigned numbers. A modified booth multiplier has a lot of partial product decrease. Floating point numbers offer increased precision and a larger range of values compared with the fixed point representation. When compared to a modified Booth MAC unit, the floating point MAC Unit performs higher on the accuracy concept. Accuracy is less critical for image and video processing applications. In order to compare approximation multiplier-based accurate adder MAC units, a variety of approximate adders are constructed, and single approximation-based adders are chosen. When compared to an accurate adder-based approximate MAC unit, an approximate adder-based approximate MAC unit consumes 22% less area and, as a result, uses less power. These blocks are implemented on an ARTIX 7 FPGA board, and various factors like area and power are compared.