Design and analysis of an approximate 2D convolver

This paper proposes a two-dimensional (2D) convolver in which both approximate circuit- and algorithm-level techniques are utilized in the design. Truncation is used as circuit techniques, while bit-width reduction is utilized at the algorithm level. These different techniques are related to the configuration of the convolver by which its operation can be configured to meet different and often contrasting figures of merit. Circuit-level simulation (using HSPICE) and an extensive evaluation of different error metrics, generic metrics such as the mean error distance (MED) and the peak signal noise ratio (PSNR) for image convolution are performed. A detailed error analysis is also presented to substantiate the simulation results. Convolution for image processing (Gaussian smoothing) is treated in detail to show the effectiveness of the proposed approach. The design, the analysis and the simulation results show that the approximate techniques utilized in the inexact convolver can operate in synergy.