Energy-efficient and fast IMPLY-based approximate full adder applying NAND gates for image processing

Today's processing applications require frequent and energy-consuming data transfers between memory and processing elements. Reducing the dimensions of Metal Oxide Semiconductor Field Effect Transistors (MOSFETs) brings essential problems. Therefore, applying emerging technologies should be considered as a solution. As a crossbar array compatible technology, memristors can play the computing and memory device roles. Therefore, it is suitable for In-Memory Computing (IMC). The issue of changing the computation method to achieve higher efficiency can also be considered. Today, approximate computing is of great interest for error-tolerant applications. These two concepts are applied, and an approximate serial full adder based on the Material Implication (IMPLY) method is introduced. The proposed Serial Approximate Full Adder using NAND gates (SAFAN) reduces the number of steps and energy consumption by up to 30% and 26% compared to approximate ones, 70% and 68% compared to exact ones, respectively. The error analysis results show that the accuracy decreases to a reasonable level by applying the SAFAN to the least significant bits (LSBs) of an approximate 8-bit adder. Image analysis criteria evaluated the functionality of the SAFAN in image processing applications, and the results showed the suitable performance of the SAFAN in error-tolerant applications.