Robust Computer Vision Under Low-Light and Low-Bandwidth Conditions

Lucas M. Anderson

Robust Computer Vision Under Low-Light and Low-Bandwidth Conditions

Author(s)	Dr. Lucas M. Anderson
Country	United States
Abstract	Computer-vision systems deployed in surveillance, intelligent transportation, remote monitoring, robotics, autonomous platforms, environmental sensing, and emergency-response environments frequently operate under conditions that differ substantially from conventional vision benchmarks. Two particularly important constraints are inadequate illumination and limited communication bandwidth. Low-light imaging reduces signal-to-noise ratio, contrast, color fidelity, edge visibility, and recoverable object detail, while bandwidth limitations force aggressive compression, reduced spatial or temporal resolution, selective frame transmission, or partitioned edge-cloud inference. ExDark contains 7,363 low-light images spanning ten illumination conditions and twelve object categories, while the DarkVision benchmark further demonstrates the importance of evaluating machine perception across multiple illumination levels and camera systems rather than assuming that conventional daylight performance generalizes to photon-limited environments. This paper proposes an integrated methodological framework for robust computer vision under simultaneous low-light and low-bandwidth constraints. The framework combines illumination-aware sensing, noise-sensitive enhancement, task-oriented feature preservation, adaptive compression, edge-cloud split inference, confidence monitoring, and degradation-aware fallback policies. Existing research provides evidence that lightweight low-light enhancement can improve dark-scene representations and that feature compression optimized directly for downstream detection can outperform conventional image-compression approaches at constrained transmission rates. A transparent simulation compares four pipeline strategies: a naïve baseline, low-light enhancement followed by conventional compression, task-aware feature compression, and a joint low-light/bandwidth-aware pipeline. Across simulated stress conditions, the mean robustness score increases from 50.83 for the baseline to 79.17 for the integrated strategy. These scores are explicitly illustrative and do not represent measurements from a real deployed system. The study argues that robustness should be optimized jointly across sensing, enhancement, representation, transmission, inference, and uncertainty management rather than treating illumination restoration and bandwidth reduction as independent engineering problems.
Keywords	low-light computer vision, bandwidth-constrained inference, robust object detection, image enhancement, edge computing, feature compression, split computing, task-aware compression, visual perception, reliable artificial intelligence
Field	Engineering
Published In	Volume 3, Issue 2, March-April 2022
Published On	2022-04-26

View / Download PDF File

About AJAMIR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+1-872-302-3851	editor@ajamir.com

American Journal of Advanced Multidisciplinary Innovation and Research

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Robust Computer Vision Under Low-Light and Low-Bandwidth Conditions

Share this