Unified-EGformer: Exposure Guided Lightweight Transformer for Mixed-Exposure Image Enhancement

Eashan Adhikarla, Kai Zhang, Rosaura G. VidalMata, Manjushree Aithal, Nikhil Ambha Madhusudhana, John Nicholson, Lichao Sun, and Brian D. Davison

Full paper (16 pages plus 6 pages supplementary)
Springer published version: https://doi.org/10.1007/978-3-031-78110-0_17
Local copy: PDF

Abstract
Despite recent strides made by AI in image processing, the issue of mixed exposure, pivotal in many real-world scenarios like surveillance and photography, remains a challenge. Traditional image enhancement techniques and current transformer models are limited with primary focus on either overexposure or underexposure. To bridge this gap, we introduce the Unified-Exposure Guided Transformer (Unified-EGformer). Our proposed solution is built upon advanced transformer architectures, equipped with local pixel-level refinement and global refinement blocks for color correction and image-wide adjustments. We employ a guided attention mechanism to precisely identify exposure-compromised regions, ensuring its adaptability across various real-world conditions. U-EGformer, with a lightweight design featuring a memory footprint (peak memory) of only 1134 MB (0.1 Million parameters) and an inference time of 95 ms (over 9x faster than typical existing implementations, is a viable choice for real-time applications such as surveillance and autonomous navigation. Additionally, our model is highly generalizable, requiring minimal fine-tuning to handle multiple tasks and datasets with a single architecture.

In Proceedings of the 27th International Conference on Pattern Recognition (ICPR 2024), pages 260-275, LNCS 15329, Kolkata, India, December 2024. Also arXiv Preprint: arXiv:2407.13170.

Back to Brian Davison's publications


Last modified: 1 December 2024
Brian D. Davison