Learning Exposure Correction in Dynamic Scenes

Jin Liu, Bo Wang, Chuanming Wang, Huiyuan Fu, Huadong Ma

Abstract: Exposure correction aims to enhance visual data suffering from improper exposures, which can greatly improve satisfactory visual effects. However, previous methods mainly focus on the image modality, and the video counterpart is less explored in the literature. Directly applying prior image-based methods to videos results in temporal incoherence with low visual quality. Through thorough investigation, we find that the development of relevant communities is limited by the absence of a benchmark dataset. Therefore, in this paper, we construct the first real-world paired video dataset, including both underexposure and overexposure dynamic scenes. To achieve spatial alignment, we utilize two DSLR cameras and a beam splitter to simultaneously capture improper and normal exposure videos. Additionally, we propose an end-to-end Video Exposure Correction Network (VECNet), in which a dual-stream module is designed to deal with both underexposure and overexposure factors, enhancing the illumination based on Retinex theory. Experimental results based on various metrics and user studies demonstrate the significance of our dataset and the effectiveness of our method. The code and dataset will be available soon.

Primary Subject Area: [Content] Media Interpretation

Relevance To Conference:

Video is one of the most mainstream and rich forms of multimedia expression. Low-quality videos with abnormal exposure often affect users' viewing experiences. Our paper aims to introduce a new video exposure correction method to achieve low-quality video enhancement and effectively improve the user experience.
Furthermore, exposure correction in low-level vision enhances the quality and usability of video data in various multimedia applications, contributing to multimedia processing. Video exposure correction ensures consistency and clarity of visual content, which is crucial for tasks such as object recognition, scene understanding, and tracking in multimedia systems. Moreover, exposure-corrected videos can provide better inputs for subsequent advanced visual tasks, such as semantic segmentation or action recognition, thus improving the performance and accuracy of multimodal processing workflows.
Therefore, video exposure correction in low-level vision plays a fundamental role in improving the user experience and enhancing the quality and usability of visual data, ultimately benefiting multimedia and multimodal processing applications.

Submission Number: 299