Bandwidth-Efficient Video Compression for Web Streaming Using Camera Localization

Majority of today's web traffic consists of highresolution videos for streaming, conferencing, and surveillance. Transmitting raw video over the web demands high bandwidth, creating a pressing need for effective compression. In this paper, we focus on compressing videos captured by moving agentssuch as drones or vehicles-used in IoT-based surveillance systems. Our method exploits 3D scene geometry and camera poses to reduce redundancy while preserving quality. Unlike approaches that rely on Structure-from-Motion, we employ an efficient Simultaneous Localization & Mapping (SLAM) technique for camera localization that can run on typical IoT devices with limited computational power. We analyze the application of SLAM for video compression and tackle the practical challenges inherent to SLAM systems. Simulation results demonstrate significant compression gains and a notable reduction in computational complexity compared to stare-of-the-art SfM methods. This makes the proposed approach highly suitable for video streaming in modern web communications and IoT applications.