A Real-time High Dynamic Range Vision System with Tone Mapping for Automotive Applications

Automated and autonomously driving vehicles capture information from the environment using vision sensors (amongst others) and process it in order to detect lane markings, pedestrians, other vehicles, or to read traffic signs. A crucial issue arises when visual sensors are used in real-world situations: the observed scenes are subject to a wide range of lighting conditions. This difficulty adds to the computational complexity needed to reliably process visual information at low-latencies. We propose a system based on a programmable vision-chip that can process and output frames in real-time in a wide range of lighting conditions and intra-scene light variations using a novel parallel High Dynamic Range (HDR) tone mapping algorithm.