Journal Search Engine
The Journal of The Korea Institute of Intelligent Transport Systems Vol.24 No.6 pp.294-308
저조도 환경 대응 차량용 경량 뉴럴 ISP 구현
Implementation of a Lightweight Neural ISP for Low-Light Automotive Applications
Abstract
This paper proposes a design and optimization method for a neural image signal processor (Neural ISP) to enhance the real-time image processing performance in automotive environments. An HDR RAW dataset was constructed using an in-vehicle camera sensor under five lighting and weather conditions—day, evening, snow, night, and night lights—and sRGB targets were generated through a traditional pipeline-based baseline ISP with saturation correction. The network architecture was based on a lightweight MW-ISPNet to ensure real-time performance, while a hue hint was introduced to improve color stability under low-light conditions. In addition, a preprocessing layer was implemented to simplify the pipeline from sensor input to Neural ISP and minimize latency. The proposed model achieved superior visual quality compared to conventional hardware ISPs in low-light environments, highlighting its potential for real-time applications in autonomous driving and intelligent transportation systems.
