Comparison of Approximate Computing with Sobel Edge Detection

Chung, Yunchul; Kim, Young‐Min

doi:10.5573/ieiespc.2021.10.4.355

Cited by 3 publications

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Also, because of the data volume and the calculation's complexity, digital systems for multimedia applications are incorporate with deep learning model which require more circuit area and power. One of the simplest ways to reduce computational complexity and energy consumption is by using approximate computing techniques [3]. Efficiency and energy usage are key considerations in the design of digital circuits.…”

Section: Introductionmentioning

confidence: 99%

Design and Analysis of Multiplexer based Approximate Adder for Low Power Applications

Eamani

Kumar

2023

IJRITCC

View full text Add to dashboard Cite

Low power consumption is crucial for error-acceptable multimedia devices, with picture compression approaches leveraging various digital processing architectures and algorithms. Humans can assemble useful information from partially inaccurate outputs in many multimedia applications. As a result, producing exact outputs is not required. The demand for an exact outcome is fading because new innovative systems are forgiving of faults. In the domain where error-tolerance is accepted, approximate computing is a new paradigm that relaxes the requirement for an accurate modeling while offering power, time, and delay benefits. Adders are an essential arithmetic module for regulating power and memory usage in digital systems. The recent implementation and use of approximate adders have been supported by trade-off characteristics such as delay, lower power consumption. This study examines the delay and power consumption of conventional and approximate adders. Also, a simple, fast, and power-efficient multiplexer-based approximate adder is proposed, and its performance outperforms the adders compared with existing adders. The proposed adder can be utilized in error-tolerant and various digital signal processing applications where exact results are not required. The proposed and existing adders are designed using EDA software for the performance calculations. With a delay of 81 pS, the proposed adder circuit reduces power consumption compared to the exact one. The experiment shows that the designed approximate adder can be used to implement circuits for image processing systems because it has a smaller delay and uses less energy.

show abstract

Section: Introductionmentioning

confidence: 99%