ES-COINA: A novel energy scalable quality-aware color interpolation architecture

Abstract

Energy-efficiency is the critical requirement for the battery operated portable devices exhibiting compute-intensive image/video processing applications. The color interpolation is the most energy consuming operation used within these devices to reconstruct the full resolution color image. In this paper, three approximate color interpolation architectures are proposed by exploiting the relative significance of the computations. These architectures provide high energy efficiency at the cost of small degradation in quality. Further, novel energy scalable color interpolation algorithm and its architecture (ES-COINA) are proposed that provide improved quality-energy trade-off. The efficacy of the proposed architectures is evaluated using various quality metrics by simulating them with different benchmark images from Kodak dataset. The proposed architectures are implemented on Artix-7 FPGA board and compared over the existing. Further, the proposed and existing architectures are implemented on Cadence RTL Compiler and synthesized with 180nm technology file. The simulation results show that the proposed approximate architectures provide 88%, 71.3%, and 46% reduced energy consumption over the best-known color interpolation architecture with a little degradation in quality. Further, the proposed ES-COINA reduces the energy consumption by 21% which can be extended to 92.6% when switched from high quality to high energy-efficient mode of operation.