Self-Gated Memory Recurrent Network for Efficient Scalable HDR Deghosting

TL;DR

This paper introduces a novel Self-Gated Memory recurrent network for HDR deghosting that outperforms existing methods, is scalable to variable-length sequences, and is computationally efficient.

Contribution

The paper proposes a new Self-Gated Memory cell architecture and a bidirectional recurrent network for HDR deghosting, achieving state-of-the-art results and scalability without re-training.

Findings

Outperforms existing HDR deghosting methods on three datasets.

Achieves scalability to variable-length input sequences.

Faster and uses fewer parameters than standard LSTM-based approaches.

Abstract

We propose a novel recurrent network-based HDR deghosting method for fusing arbitrary length dynamic sequences. The proposed method uses convolutional and recurrent architectures to generate visually pleasing, ghosting-free HDR images. We introduce a new recurrent cell architecture, namely Self-Gated Memory (SGM) cell, that outperforms the standard LSTM cell while containing fewer parameters and having faster running times. In the SGM cell, the information flow through a gate is controlled by multiplying the gate's output by a function of itself. Additionally, we use two SGM cells in a bidirectional setting to improve output quality. The proposed approach achieves state-of-the-art performance compared to existing HDR deghosting methods quantitatively across three publicly available datasets while simultaneously achieving scalability to fuse variable-length input sequence without necessitating re-training. Through extensive ablations, we demonstrate the importance of individual components in our proposed approach. The code is available at https://val.cds.iisc.ac.in/HDR/HDRRNN/index.html.