Digital image processing in professional applications places ever-higher demands in terms of compute power and power consumption so that FPGA devices reach their limits. Approximate Computing is an approach to tackle this challenge. It refers to a set of methods to perform calculations not exactly but only approximately. As a benefit, fewer resources are used in the FPGA, more functions can be implemented in FPGA devices, and the energy efficiency of the calculations is improved. However, exploring the design space for multi-dimensional Pareto optimization requires long simulation runs in order to find a trade-off between image quality and FPGA resources/power.
The paper entitled "Region of Interest-Based Parameter Optimization for Approximate Image Processing on FPGAs" addresses this problem and has recently been published in the International Journal of Networking and Computing (https://www.jstage.jst.go.jp/article/ijnc/11/2/11_438/_article).
It proposes and evaluates a method for reducing the computational load of the design space exploration and thus to converge faster to the Pareto front. The authors are Manu Manuel, Arne Kreddig, Simon Conrady, Nguyen Anh Vu Doan, and Walter Stechele.
A recent conference paper entitled "Region of Interest Based Non-dominated Sorting Genetic Algorithm-II: An Invite and Conquer Approach" has a broader focus on benchmark applications not limited to approximate computing. It has been accepted for presentation at the Genetic and Evolutionary Computation Conference in Boston, July 9-13, 2022 (https://gecco-2022.sigevo.org/HomePage). The authors of this paper are are Manu Manuel, Benjamin Hien, Simon Conrady, Arne Kreddig, Nguyen Anh Vu Doan, and Walter Stechele.
Congratulation to the authors.