Learning a Reinforced Agent for Flexible Exposure Bracketing Selection

Zhouxia Wang, Jiawei Zhang, Mude Lin, Jiong Wang, Ping Luo, Jimmy Ren; The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 1820-1828

Abstract


Automatically selecting exposure bracketing (images exposed differently) is important to obtain a high dynamic range image by using multi-exposure fusion. Unlike previous methods that have many restrictions such as requiring camera response function, sensor noise model, and a stream of preview images with different exposures (not accessible in some scenarios e.g. mobile applications), we propose a novel deep neural network to automatically select exposure bracketing, named EBSNet, which is sufficiently flexible without having the above restrictions. EBSNet is formulated as a reinforced agent that is trained by maximizing rewards provided by a multi-exposure fusion network (MEFNet). By utilizing the illumination and semantic information extracted from just a single auto-exposure preview image, EBSNet enables to select an optimal exposure bracketing for multi-exposure fusion. EBSNet and MEFNet can be jointly trained to produce favorable results against recent state-of-the-art approaches. To facilitate future research, we provide a new benchmark dataset for multi-exposure selection and fusion.

Related Material


[pdf] [supp]
[bibtex]
@InProceedings{Wang_2020_CVPR,
author = {Wang, Zhouxia and Zhang, Jiawei and Lin, Mude and Wang, Jiong and Luo, Ping and Ren, Jimmy},
title = {Learning a Reinforced Agent for Flexible Exposure Bracketing Selection},
booktitle = {The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}