A 26.4mW, 18.6MS/s Image Reconstruction Processor for IoT Compressive Sensing

This work presents the first energy-efficient, real-time image reconstruction processor for IoT compressive sensing. The chip implemented in 40 nm CMOS dissipates only 26.4 mW with a throughput of 60 fps for VGA images, delivering $5.4 \times 10^{6} \times$ and $6.5 \times 10^{4} \times$ higher energy and area efficiency than a high-end CPU. It supports an equivalent throughput of 18.6 MS/s for 2 D images, achieving $27.2 \times$ and $19.5 \times$ improvements in energy and area efficiency than state-of-the-art 1D compressive sensing reconstruction processors.