DeNIM: Deterministic Neural Illuminant Mapping for Efficient Auto-White Balance Correction

Deterministic Neural Illuminant Mapping for Efficient Auto-White Balance Correction
Furkan Kınlı, Doğa Yılmaz, Barış Özcan, Furkan Kıraç
Accepted to RCV2023 at ICCV2023

Abstract: Auto-white balance (AWB) correction is a critical operation in image signal processors (ISPs) for accurate and consistent color correction across various illumination scenarios. This paper presents a novel and efficient AWB correction method that achieves at least 35 times faster processing with equivalent or superior performance on high-resolution images for the current state-of-the-art methods. Inspired by deterministic color style transfer, our approach introduces deterministic illumination color mapping, leveraging learnable projection matrices for both canonical illumination form and AWB-corrected output. It involves feeding high-resolution images and corresponding latent representations into a mapping module to derive a canonical form, followed by another mapping module that maps the pixel values to those for the corrected version. This strategy is designed as resolution-agnostic and also enables seamless integration of any pre-trained AWB network as the backbone. Experimental results confirm the effectiveness of our approach, revealing significant performance improvements and reduced time complexity compared to state-of-the-art methods. Our method provides an efficient deep learning-based AWB correction solution, promising real-time, high- quality color correction for digital imaging applications.

Description

The official implementation of the paper titled "Deterministic Neural Illuminant Mapping for Efficient Auto-White Balance Correction". We propose a novel and efficient strategy for AWB correction, which learns deterministic color mappings for both canonical illumination and AWB-corrected forms with the help of learnable projection matrices.

Requirements

To install requirements:

pip install -r requirements.txt

Architecture

Updates

8/8/2023: Release of the code

6/8/2023 Accepted to RCV2023 in conjunction with ICCV2023

26/7/2023: Submission of the paper to RCV2023 at ICCV2023

Training

To train DeNIM from the scratch in the paper, run this command:

python main.py --cfg configs/train_<backbone_type>_<patch_size>_<num_channels>.yaml

• Available backbones: "style_wb", "mixed_wb"
• Available patch sizes: 64, 128
• Available number of channels for different WB settings: 9, 15

Pre-trained weights

To prepare pre-trained weights for both backbones and DeNIM, download files from link, then:

mv weights.zip DeNIM/
cd DeNIM
unzip weights.zip

Evaluation

To evaluate DeNIM on Cube+ dataset, run:

python main.py --cfg configs/test_<backbone_type>_<patch_size>_<num_channels>.yaml --is_train False

Citation

TDB

Contacts

Please feel free to open an issue or to send an e-mail to [email protected]