The phrase “garbage in, garbage out” is applicable to machine and human vision. If we can improve the quality of image data at the source by removing noise, this will have far-reaching impacts, such as improved accuracy later in the pipeline, particularly in the challenging conditions of low light or high dynamic range. In this talk, we will present our new, AI-based approach for removing image noise and preserving image quality in real time, and share the latest results demonstrating how it performs with commonly used image sensors. We’ll also show that we are able to achieve these results with very low compute resource requirements, making our approach suitable for battery-powered devices at the edge.