信号处理2025，Vol.41Issue(2)：224-240,17.DOI:10.12466/xhcl.2025.02.003
基于RAW数据的图像处理及应用
Image Processing and Applications Based on RAW Sensor Data
摘要
Abstract
In the development of computer-vision technology,researchers have primarily focused on processing stan-dard RGB images preprocessed using an image signal processor(ISP).These images are small as well as convenient for use and network transmission;thus,they are widely applied in many typical application scenarios.However,under low-light conditions or extreme imaging environments,these compressed and processed images typically exhibit irreversible degradations such as blurring and quantization,thus resulting in detail loss and limited performance.Hence,increasing attention has been focused on the direct processing of RAW images output by camera sensors.RAW images,which do not require complex ISP processing,exhibit characteristics such as linear response,large bit depth,and lossless com-pression,which allow them to retain more original sensory information.These features render RAW images highly flex-ible and capable in low-light,high dynamic range,and complex visual scenarios.In recent years,RAW-image process-ing technologies have developed significantly,with their applications extending from high-quality image and video ac-quisition,denoising,and enhancement to computer-vision tasks such as object recognition and scene understanding.Compared with conventional RGB images,RAW images preserve details more accurately and significantly enhance the accuracy and robustness of visual tasks under specific conditions.Furthermore,the development of deep-learning tech-niques has enabled end-to-end models based on RAW data,where the original signal information in images are used to enhance visual processing performance.This paper systematically reviews the most recent progress in RAW-image pro-cessing technologies and discusses their applications in various computer-vision fields.Additionally,future development trends are forecasted,particularly the potential of using RAW-image data in more complex scenarios,thus providing valuable reference and insights for researchers and practitioners in the field.关键词
RAW数据/图像处理/计算机视觉/图像增强/视觉技术应用Key words
RAW data/image processing/computer vision/image enhancement/vision applications分类
信息技术与安全科学引用本文复制引用
邹云昊,李和松,聂婧,姜雨彤,付莹..基于RAW数据的图像处理及应用[J].信号处理,2025,41(2):224-240,17.基金项目
国家自然科学基金(62331006,62171038) （62331006,62171038）
中央高校基本科研业务费资助The National Natural Science Foundation of China(62331006,62171038) （62331006,62171038）
The Fundamental Research Funds for the Central Universities （）
