Abstract

This paper describes a dual-stream video delivery system for live broadcasting that reduces transmitted bandwidth by encoding only the region of active motion at full resolution. In typical live sports broadcasts, the area of viewer interest occupies only 15-20% of the video frame, while the surrounding background consumes bandwidth at uniform quality. The described system repurposes raw motion vectors generated during the AOMedia Video 1 (AV1) encoding pass to compute an action centroid coordinate, bypassing neural network-based analysis.

The encoder splits the broadcast into two concurrent layers: a full-frame background aggressively compressed to low resolution and a bounding-box action crop encoded at maximum fidelity. A time-synced coordinate matrix maps the crop position onto the background. On the client device, native mobile graphics APIs composite the layers while an alpha-blending shader applies perimeter feathering to eliminate visible seam artifacts. This system achieves approximately 75% reduction in Content Delivery Network (CDN) transit bandwidth while preserving perceived 4K fidelity in the focal area and supports instant vertical aspect ratio translation without server-side transcoding.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS