Deep Learning-Based Patch-Wise Illumination Estimation for Enhanced Multi-Exposure Fusion

This article suggests a unique technique for multi-exposure fusion using convolutional neural networks (CNNs) for patch-wise illumination estimates. Multi-exposure fusion is a crucial component of enhancing image quality, particularly in circumstances with erratic lighting. Our proposed approach makes use of CNNs' capability to anticipate light levels inside specific image patches in order to accurately change exposure levels. We look at the theoretical foundations of our approach, emphasising the advantages of patch-wise estimation in capturing intricate lighting details. Additionally, we present experimental results demonstrating enhanced dynamic range expansion and image detail preservation, demonstrating that our methodology is more effective than conventional fusion methods. This study advances the state-of-the-art in multi-exposure fusion while also opening up new prospects for computational photography, surveillance, and computer vision applications.