Saturation-Enhanced Dynamic Range Network for HDR-like Image Enhancement

The prevalence of the challenge in real-time object detection, particularly under varying lighting conditions, underscores a significant limitation of current HDR cameras due to the reliance on multiple exposures, which is impractical in dynamic scenes or with moving objects. This research introduces a novel network, SatHDR-Net, capable of generating HDR-like images from a single LDR input. It utilizes three innovative loss functions—Kullback-Leibler Divergence Loss for color fidelity, Dynamic Range Loss for histogram optimization, and a loss function targeting HSV difference to maintain color saturation—effectively enhancing image quality while preserving original details. Simulation results demonstrate the model efficiency, achieving superior performance metrics such as the lowest recorded inference time of 2.8404 seconds .This emphasizes the network potential in applications requiring rapid processing and high-quality image enhancement.