Lightweight HDR Deghosting Network with Parallel Vision Transformer

High dynamic range (HDR) deghosting networks aim to generate ghost-free HDR images from multiple low dynamic range (LDR) images. However, previous models have been hampered by redundant parameters and lengthy inference time, thereby constraining their applicability on edge devices. In this paper, we propose a lightweight model that includes Parallel Vision Transformer structures for ghost-free HDR imaging. The results demonstrate that our model can produce high-quality HDR images with fewer parameters and greater efficiency compared to state-of-the-art methods.