Learning to See in the Dark

Processing low-light images with minimal noise

Researchers with the University of Illinois Urbana–Champaign and Intel have developed a deep neural network that brightens ultra-low light images without adding noise and other artefacts.  The network was trained using 5,094 raw short-exposure low-light and long-exposure image pairs—the end result is a system that automatically brightens images at a much higher quality than traditional processing options.

Traditional methods to process low light images result typically in high levels of noise that is absent using the machine learning process:

Learning to see in the dark

These almost unbelievable examples of machine learning are revealed in the video below:

Paper: HERE

Project Page: HERE

Leave a Reply