Efficient Utilization of Imprecise Blocks for Hardware Implementation of a Gaussian Filter

Due to great advantages of the imprecise computational blocks in implementation of imprecision tolerant applications, a wide range of different imprecise structures as well as some of their new applications are introduced in recent years. However, due to inherent differences between imprecise and precise components, their utilization approaches should also be customized to gain the best performance. In this paper, a novel efficient approach for utilization of imprecise computational blocks is presented and applied to a Gaussian image filter hardware. This approach offers different imprecise implementations of the Gaussian filter which improve the filter area and delay up to 50% and 53% respectively while does not degrade the output quality below the excellent threshold.