# SMAE: Few-shot Learning for HDR Deghosting with Saturation-Aware Masked Autoencoders

arXiv: https://arxiv.org/abs/2304.06914 · HTML: https://mlsanity.com/paper/2304.06914/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- Builds on 14 of its 41 references that are in the map.
- First posted on arXiv 14 Apr 2023.

## Builds on (most cited first, 14 of 14)

- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks](https://mlsanity.com/paper/1703.03400/) (2017)
- [SwinIR: Image Restoration Using Swin Transformer](https://mlsanity.com/paper/2108.10257/) (2021)
- [EDVR: Video Restoration with Enhanced Deformable Convolutional Networks](https://mlsanity.com/paper/1905.02716/) (2019)
- [Key-Value Memory Networks for Directly Reading Documents](https://mlsanity.com/paper/1606.03126/) (2016)
- [Bayesian Model-Agnostic Meta-Learning](https://mlsanity.com/paper/1806.03836/) (2018)
- [Attention-guided Network for Ghost-free High Dynamic Range Imaging](https://mlsanity.com/paper/1904.10293/) (2019)
- [Deep High Dynamic Range Imaging with Large Foreground Motions](https://mlsanity.com/paper/1711.08937/) (2017)
- [Learning feed-forward one-shot learners](https://mlsanity.com/paper/1606.05233/) (2016)
- [HDR-GAN: HDR Image Reconstruction from Multi-Exposed LDR Images with Large Motions](https://mlsanity.com/paper/2007.01628/) (2020)
- [ADNet: Attention-guided Deformable Convolutional Network for High Dynamic Range Imaging](https://mlsanity.com/paper/2105.10697/) (2021)
- [One-Shot Unsupervised Cross Domain Translation](https://mlsanity.com/paper/1806.06029/) (2018)
- [Progressive and Selective Fusion Network for High Dynamic Range Imaging](https://mlsanity.com/paper/2108.08585/) (2021)
- [Attentive Recurrent Comparators](https://mlsanity.com/paper/1703.00767/) (2017)

## Built on by (most cited first, 11 of 11)

- [Masked Image Modeling: A Survey](https://mlsanity.com/paper/2408.06687/) (2024)
- [LEDiff: Latent Exposure Diffusion for HDR Generation](https://mlsanity.com/paper/2412.14456/) (2024)
- [Cinematic Gaussians: Real-Time HDR Radiance Fields with Depth of Field](https://mlsanity.com/paper/2406.07329/) (2024)
- [Self-Supervised High Dynamic Range Imaging with Multi-Exposure Images in Dynamic Scenes](https://mlsanity.com/paper/2310.01840/) (2023)
- [CRNet: A Detail-Preserving Network for Unified Image Restoration and Enhancement Task](https://mlsanity.com/paper/2404.14132/) (2024)
- [Single-Image HDR Reconstruction Assisted Ghost Suppression and Detail Preservation Network for Multi-Exposure HDR Imaging](https://mlsanity.com/paper/2403.04228/) (2024)
- [PASTA: Towards Flexible and Efficient HDR Imaging Via Progressively Aggregated Spatio-Temporal Alignment](https://mlsanity.com/paper/2403.10376/) (2024)
- [Few-Shot Vision-Language Action-Incremental Policy Learning](https://mlsanity.com/paper/2504.15517/) (2025)
- [DiffHDR: Re-Exposing LDR Videos with Video Diffusion Models](https://mlsanity.com/paper/2604.06161/) (2026)
- [Cross-Modal Few-Shot Learning: a Generative Transfer Learning Framework](https://mlsanity.com/paper/2410.10663/) (2024)
- [InstantHDR: Single-forward Gaussian Splatting Initialization for HDR 3D Reconstruction](https://mlsanity.com/paper/2603.11298/) (2026)

## Abstract (arXiv)

Generating a high-quality High Dynamic Range (HDR) image from dynamic scenes has recently been extensively studied by exploiting Deep Neural Networks (DNNs). Most DNNs-based methods require a large amount of training data with ground truth, requiring tedious and time-consuming work. Few-shot HDR imaging aims to generate satisfactory images with limited data. However, it is difficult for modern DNNs to avoid overfitting when trained on only a few images. In this work, we propose a novel semi-supervised approach to realize few-shot HDR imaging via two stages of training, called SSHDR. Unlikely previous methods, directly recovering content and removing ghosts simultaneously, which is hard to achieve optimum, we first generate content of saturated regions with a self-supervised mechanism and then address ghosts via an iterative semi-supervised learning framework. Concretely, considering that saturated regions can be regarded as masking Low Dynamic Range (LDR) input regions, we design a Saturated Mask AutoEncoder (SMAE) to learn a robust feature representation and reconstruct a non-saturated HDR image. We also propose an adaptive pseudo-label selection strategy to pick high-quality HDR pseudo-labels in the second stage to avoid the effect of mislabeled samples. Experiments demonstrate that SSHDR outperforms state-of-the-art methods quantitatively and qualitatively within and across different datasets, achieving appealing HDR visualization with few labeled samples.