Invertible Tone Mapping with Selectable Styles

Zhuming Zhang,Menghan Xia,Xueting Liu,Chengze Li,Tien-Tsin Wong

DOI: https://doi.org/10.48550/arXiv.2110.04491

2021-10-09

Abstract:Although digital cameras can acquire high-dynamic range (HDR) images, the captured HDR information are mostly quantized to low-dynamic range (LDR) images for display compatibility and compact storage. In this paper, we propose an invertible tone mapping method that converts the multi-exposure HDR to a true LDR (8-bit per color channel) and reserves the capability to accurately restore the original HDR from this {\em invertible LDR}. Our invertible LDR can mimic the appearance of a user-selected tone mapping style. It can be shared over any existing social network platforms that may re-encode or format-convert the uploaded images, without much hurting the accuracy of the restored HDR counterpart. To achieve this, we regard the tone mapping and the restoration as coupled processes, and formulate them as an encoding-and-decoding problem through convolutional neural networks. Particularly, our model supports pluggable style modulators, each of which bakes a specific tone mapping style, and thus favors the application flexibility. Our method is evaluated with a rich variety of HDR images and multiple tone mapping operators, which shows the superiority over relevant state-of-the-art methods. Also, we conduct ablation study to justify our design and discuss the robustness and generality toward real applications.

Image and Video Processing,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

### What problem does this paper attempt to solve?

This paper aims to solve the problem of information loss when high - dynamic - range (HDR) images are converted into low - dynamic - range (LDR) images. Specifically, modern digital cameras can capture high - precision HDR images, but these images are usually quantized into LDR images for display and storage. This quantization process will lead to permanent loss of visual details, such as over - exposed skies, etc.

To overcome this problem, the paper proposes an **Invertible Tone Mapping with Selectable Styles**, which can convert HDR images into standard 8 - bit LDR images and retain the ability to accurately recover the original HDR images from these LDR images. In addition, the generated LDR images can also imitate the tone - mapping styles selected by users, allowing users to perform different tone - mapping processes as needed.

### Main objectives

1. **Accurately recover HDR images**: Ensure that the original HDR images can be accurately recovered from the generated LDR images.
2. **Support multiple tone - mapping styles**: The generated LDR images can imitate the tone - mapping styles selected by users, thus providing flexibility.
3. **Be compatible with existing social platforms**: The generated LDR images can be shared on existing social platforms, and even after re - encoding or format conversion, it will not significantly affect the accuracy of HDR recovery.

### Technical implementation

To achieve the above objectives, the paper proposes the following technical solutions:

- **Encoding - decoding framework**: Consider tone - mapping and recovery as a whole, and implement the encoding and decoding processes through a convolutional neural network (CNN).
- **Two - branch network architecture**: Use global and local receptive fields to comprehensively capture image information to support controllable tone - mapping styles.
- **Plug - in style regulators**: Support the flexible insertion of multiple tone - mapping styles, and each style regulator can bake a specific tone - mapping style.
- **Loss function design**: Introduce invertibility loss and style loss to ensure that the recovered HDR images are close to the original HDR images and that the generated LDR images have the desired visual effects.

Through these technical means, the invertible tone - mapping method proposed in the paper can not only retain HDR information in LDR images, but also support multiple tone - mapping styles, thus providing a new solution for the processing and distribution of HDR images.