高动态范围图像生成方法研究进展OA

High dynamic range(HDR)images effectively represent the complex luminance distributions found in natural scenes,leading to significant advantages in highlights retention,shadow detail recovery,and overall visual fidelity.Conse-quently,HDR imaging has become a prominent research direction within the fields of computer vision and computational imaging.With the advancement of deep learning technologies and the growing demand for multi-scene perceptual applica-tions,HDR image generation techniques exhibit vast potential across various scenarios,including mobile photography,remote sensing imaging,intelligent surveillance,and industrial inspection.However,existing HDR image generation methods still face numerous challenges regarding imaging quality,detail reconstruction,generalization capabilities,and evaluation criteria.Furthermore,relevant research findings remain fragmented,lacking a systematic review and summary.To address this situation,this paper provides a comprehensive overview and analysis of the research progress in HDR image generation.Initially,based on different technical approaches to HDR image generation,existing methods are systematically categorized,highlighting the fundamental principles,development processes,applicable conditions,and their strengths and weaknesses.Next,commonly used performance evaluation metrics in HDR image generation tasks are compiled and analyzed,discussing the applicability and limitations of various metrics.Additionally,this paper summarizes the mainstream datasets frequently utilized in current HDR image generation research,examining their data sources,scale characteristics,scene coverage,and usage in training and testing,offering guidance for future research.Lastly,in conjunction with existing research findings and development trends,this paper anticipates potential future directions for HDR image generation technology.This paper aims to provide a systematic theoretical reference and method for HDR image generation-related research and to offer insights for subsequent algorithm design,dataset construction,and evalua-tion system enhancement.