Learning Differential Pyramid Representation for Tone Mapping
Keywords: Tone Mapping, Differential Pyramid, Image Signal Processor, High Dynamic Range, Image Retouching
Abstract: To display high dynamic range (HDR) images on low dynamic range (LDR) screens, tone mapping operations (TMO) are required to compress the dynamic range. Recently, the deep learning-based TMO methods with 3D Look-Up Table (LUT) have shown promising performance. However, these methods often fail to deliver satisfactory results in local areas, and generating image-level TMO on down-sampled low-resolution images leads to loss of details. To overcome this problem, we propose to construct a learnable differential pyramid representation network, termed DPRNet, for joint global and local tone mapping. Specifically, we construct multi-layer perceptrons to globally modulate the tones in pixel-level. Then, we propose a local 3D LUT, which generates the TMO coefficients in patch-level. To further enhance the details, we propose a learnable differential pyramid to capture multi-scale high-frequency components, coupled with an iterative mask learning strategy to refine high-frequency details. Extensive experiments demonstrate that our method significantly outperforms state-of-the-art methods, improving PSNR by 2.58 dB in the HDR+ dataset and 3.31 dB in the HDRI Haven dataset respectively compared with the second best method. In addition, our method has the best generalization ability in unsupervised video TMO. We provide an anonymous online demo at https://xxxxxx2024.github.io/DPRNet/ .
Primary Area: applications to computer vision, audio, language, and other modalities
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 5379
Loading