Illumination Estimation Challenges
Welcome to the "Night Photography" challenge part of the NTIRE workshop at CVPR 2025. The challenge seeks to advance image processing methods for night photography by addressing the complexities of rendering images captured at night, particularly this year with raw mobile phone captures, and evaluating the results based on perceptual quality and computational efficiency.

News and updates

January 24th, 2025

Motivation for this challenge

The process of capturing and processing images taken by cameras relies on onboard processing to transform raw sensor data into polished photographs, typically encoded in a standard color space such as sRGB. Night photography, however, presents unique challenges not encountered in daylight scenes. Night images often feature complex lighting conditions, including multiple visible illuminants, and are characterized by distinctive noise patterns that make conventional photo-finishing techniques for daytime photography unsuitable. Additionally, widely used image quality metrics like SSIM and LPIPS are often ineffective for evaluating the nuances of night photography. In previous editions of this challenge, participants were tasked with processing raw night scene images into visually pleasing renders, assessed subjectively using mean opinion scores. These efforts have driven significant advancements in the field of night image processing.

This year's challenge introduces a new and fundamentally different approach while retaining the use of the mean opinion score as an evaluation metric. Moving away from purely subjective aesthetics, the 2025 challenge will adopt a potentially more objective framework by leveraging paired datasets captured with a Huawei smartphone and a high-end Sony camera using a beam splitter to ensure identical perspectives. Participants will be provided with raw Huawei smartphone images as input and corresponding processed Sony camera images as the expected output. The goal is to develop algorithms that transform the raw Huawei images—characterized by higher noise levels, vignetting, and reduced detail—into outputs that convincingly resemble the high-quality processed Sony images.

Mean opinion score will be employed to evaluate submissions, ensuring that human viewers, rather than automated metrics that might be exploited, determine how convincingly the processed Huawei images align with the processed Sony images. This focus on human perception underscores the importance of developing rendering algorithms that produce results that are not only technically accurate but also perceptually closer from the target images.

By addressing the unique limitations of mobile phone raw images and the inherent complexities of night photography, the 2025 challenge seeks to advance the state of the art in image processing. The combination of objective input-output alignment and perceptual validation through mean opinion scores provides a robust framework for fostering innovation in mobile-focused night photography rendering where objectivity has a more pronounced role, but is not expected to be abused in anyway due to the final subjective oversight.

Challenge Goal and Uniqueness

This challenge tackles the complexities of nighttime photography by leveraging paired datasets of raw Huawei smartphone images and processed Sony camera images, providing a clear ground truth for evaluation. The goal is to develop algorithms that process raw Huawei images to convincingly resemble the high-quality Sony outputs, addressing a long-standing challenge in computer vision.

Night photography is vital for applications like surveillance and security and also has artistic significance in creating stunning images. By combining objective ground-truth comparisons with human perception through mean opinion scores, the challenge ensures that results are both technically accurate and visually convincing.

This unique framework bridges mobile device constraints, low-light conditions, and human-centric evaluation, advancing the state of the art in night image processing.

Challenge data

The participants in this challenge will be granted access to raw-RGB images of night scenes, which have been simultaneously captured using a Huawei smartphone sensor and a Sony professional camera sensor by means of a beam splitter. The images are encoded in 16-bit PNG files. Accompanying these images will be additional meta-data, provided in JSON files. The challenge will commence with the provision of the initial images to participants, which will be utilized for the development and testing of their algorithms. Data access will be granted upon registration, and further information can be found on the registration form at the bottom of the page. Additional images will be made available throughout the challenge, and further details can be found in the information provided on the evaluation and leaderboard. As an added resource, the organizers have also provided code for a baseline algorithm, as well as a demonstration, on GitHub.

Evaluation/Leaderboard

More information on evaluation and leadersboard formation will be given soon.

Submissions

The submission rules will be described here around the same time when the challenge data becomes available.

Timeline

* Please note that the timeline can be slightly changed, so it is advised to check the challenge web page over time.

Registered Teams

The list of registered teams will be available here.

Reporting

In order to be eligible for the prizes, the participants will be required to send code and reports about their solutions in the form of short papers during the submission. If this report is not submitted, the participants that would otherwise win a prize will be passed over.

Prizes

Winners will receive a winner certificate and will have an opportunity to submit their paper to NTIRE'2025 and participate in the common report which also will be submitted to CVPR workshop.

More information on the monerary prizes will be given soon.