English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Deep learning-based low-light image enhancement (LLIE) is a task ofleveraging deep neural networks to enhance the image illumination while keepingthe image content unchanged. From the perspective of training data, existingmethods complete the LLIE task driven by one of the following three data types:paired data, unpaired data and zero-reference data. Each type of thesedata-driven methods has its own advantages, e.g., zero-reference data-basedmethods have very low requirements on training data and can meet the humanneeds in many scenarios. In this paper, we leverage pure Gaussian noise tocomplete the LLIE task, which further reduces the requirements for trainingdata in LLIE tasks and can be used as another alternative in practical use.Specifically, we propose Noise SElf-Regression (NoiSER) without access to anytask-related data, simply learns a convolutional neural network equipped withan instance-normalization layer by taking a random noise image,$\mathcal{N}(0,\sigma^2)$ for each pixel, as both input and output for eachtraining pair, and then the low-light image is fed to the trained network forpredicting the normal-light image. Technically, an intuitive explanation forits effectiveness is as follows: 1) the self-regression reconstructs thecontrast between adjacent pixels of the input image, 2) theinstance-normalization layer may naturally remediate the overallmagnitude/lighting of the input image, and 3) the $\mathcal{N}(0,\sigma^2)$assumption for each pixel enforces the output image to follow the well-knowngray-world hypothesis when the image size is big enough. Compared to currentstate-of-the-art LLIE methods with access to different task-related data,NoiSER is highly competitive in enhancement quality, yet with a much smallermodel size, and much lower training and inference cost. Besides, NoiSER alsoexcels in mitigating overexposure and handling joint tasks.

Noise Self-Regression: A New Learning Paradigm to Enhance Low-Light Images Without Task-Related Data