Multi-exposure image acquisition for automotive high dynamic range imaging

Abstract

Image acquisition is the elementary step on the way to a performing vision-based Driver Assistance System (DAS). Any visual information lost in the capturing process is irretrievably gone. Many application approaches do not consider any influence on the camera, but take the image data as is from the capturing device. This paper starts with an overview of the most relevant solutions of high dynamic range (HDR) imaging and introduces a novel and real-time capable approach of using multiple low dynamic range (LDR) images from an automotive camera. Especially the automatic exposure and gain control for the different exposed images is in the focus. In the next step the HDR histogram linearization and image fusion to a final 8-bit HDR-image is shown. The exposure control algorithm is proven to be fast and stable. The resulting image shows both dark and bright details accurately.