Video Bandwidth Optimization via Face Detection and Adaptive Bit-Rate Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In networks with limited bandwidth, existing technologies fail to effectively optimize video bandwidth, leading to suboptimal video quality in video conferencing and other media applications.
Innovation Solution
The implementation of video analytics, such as face detection and eye gaze detection, in conjunction with adaptive bit-rate codecs and scalable video coding, to analyze and process video streams, reducing bandwidth requirements by adjusting bit rates and dropping layers based on participant presence and viewing behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If video bandwidth optimization is implemented in networks with limited bandwidth, then video quality is improved, but device complexity increases due to video analytics processing
Solution Approach 1:
The system performs video analytics processing in advance to identify faces and determine eye gaze directions before video transmission. This preliminary analysis enables the network device to pre-determine which video portions require higher bit rates, allowing optimized encoding without real-time processing delays and reducing the complexity of real-time adaptive processing.
Solution Approach 2:
The patent applies different bit rate allocations to different spatial regions of the video based on local importance. Faces and eye gaze regions receive higher bit rate allocation while other portions receive lower allocation. This local quality differentiation improves overall video quality perception while reducing total bandwidth requirements, resolving the contradiction between quality and complexity.
2Quantity of substance
If adaptive bit-rate encoding is used to optimize bandwidth, then bandwidth efficiency is improved, but loss of information increases in non-critical video portions
Solution Approach 1:
The system applies differential quality encoding where only non-critical portions of the video (areas outside faces and eye gaze regions) receive reduced bit rate allocation. Critical regions maintain high quality encoding. This selective approach improves bandwidth efficiency while minimizing information loss in perceptually important areas.
Solution Approach 2:
The video analytics module acts as an intermediary that identifies and marks critical regions (faces, eye gaze) in the video stream. This intermediary processing enables the encoder to selectively apply different quality levels without losing critical information, as the analytics module precisely delineates which regions require full quality preservation.
3Quantity of substance
If video analytics processing is performed to identify faces and eye gaze, then bandwidth allocation is optimized, but processing time increases
Solution Approach 1:
The system performs video analytics processing as a preliminary step before video transmission and encoding. By completing face detection and eye gaze analysis in advance, the system establishes the basis for bandwidth optimization without adding real-time processing delays during video delivery, thus improving bandwidth allocation efficiency while minimizing time loss.
Data Source
AI summary
In one embodiment, a method includes receiving at a network device, video from a first endpoint, analyzing the video received from the first endpoint, processing video received from a second endpoint based on the analyzed video to optimize bandwidth between the network device and the first endpoint, and transmitting the processed video to the first endpoint. An apparatus is also disclosed.