"DIMEPlus: Learning Video Exposure Correction Benchmark Dataset"

"Exposure correction aims to enhance visual data degraded by improper exposures, significantly improving perceptual quality. Although substantial progress has been achieved for single-image exposure correction, its extension to videos remains underexplored. Directly applying single-frame-based methods to videos results in temporal incoherence and compromised visual quality. Through systematic investigation, we identify that the absence of a benchmark dataset has hindered progress in this domain. To address this gap, we introduce a real-world paired video dataset covering both underexposed and overexposed dynamic scenes. The dataset comprises 36K spatially aligned frame pairs, captured using a dual-camera acquisition system with a beam splitter, enabling synchronized recording of improperly exposed and reference videos."