CAGGLE: Color-Aware Guidance with Global and Local Prompts for Exposure Correction

Daehyun Kim, Donggoo Jung, Guanghui Wang, Tae Hyun Kim

Keywords: prompt learning, exposure correction, image enhancement

TL;DR: Our model is the first prompt-based learning approach specifically designed for exposure correction.

Abstract:

In real-world exposure correction, achieving high-quality images requires addressing multi-exposure conditions and managing images containing locally varying brightness.
While recent deep learning models have improved image correction across various exposure levels, they often struggle in complex scenarios where both under- and over-exposure coexist within an image or in over-saturated areas with sparse pixel information.
In this paper, we tackle these challenges by proposing a color-aware guidance that employs a global prompt for tone adjustment and a local prompt for maintaining color consistency of the output.
To achieve this, we present a novel Prompt Interaction Module (PIM) that seamlessly integrates the global and local prompts with the input image features.
Extensive experiments on multi-exposure benchmark datasets demonstrate that our method achieves state-of-the-art performance, outperforming existing exposure correction methods.
Our approach sets a new standard in exposure correction, leveraging prompt-based learning for improved color and exposure adjustments.

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: 6153