This work presents a new analog architecture to perform image convolution for deep learning purposes in CMOS imagers in the analog domain. The architecture is focused to reduce both power dissipation ...