Year: 2025
Type: article
Abstract: This article presents a physics-based simulation that models the complete imaging pipeline from scene radiance to the final rendered image. We use this simulation to evaluate sensor designs optimized for high dynamic range (HDR) environments, such as driving through daytime tunnels or in nighttime conditions. The work makes three main contributions: 1) a synthetic, labeled dataset of HDR driving scenes with instance segmentation and depth information; 2) open-source simulation software with validated performance; and 3) a comparative analysis of two single-shot sensor designs optimized for HDR imaging. Both the dataset [(ISETHDR)] and simulation software (ISETHDRSensor) are made publicly available and can be used to evaluate sensor designs for HDR environments, such as nighttime driving scenes.
Source: IEEE Sensors Journal
Authors: Zhenyi Liu, Devesh Shah, Brian A. Wandell
Institutions: Stanford University, Ford Motor Company (United States)
Language: English
Topic: Image Enhancement Techniques
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: Ford Motor Company
OpenAlex ID: W4408519888