Videoconferencing Endpoint Image Privacy Using Monocular Depth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing videoconferencing systems fail to differentiate between areas at an endpoint that are suitable for sharing and those that are not, leading to unintended disclosure of extraneous image data.
Innovation Solution
A videoconferencing system that uses monocular depth estimation and computer vision techniques to determine regions of interest and exclusion, allowing users to select areas in three-dimensions that will not be shared with a remote endpoint, and replaces or obscures image data from excluded regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If the image capture device captures all visible areas in the room, then the completeness of the transmitted image data is improved, but extraneous or unwanted image data may be inadvertently shared with remote endpoints
Solution Approach 1:
The patent divides the captured image into multiple depth layers using monocular depth estimation, separating foreground objects (meeting participants) from background areas. This segmentation allows selective transmission of only relevant image portions while excluding extraneous data, resolving the contradiction between completeness and privacy protection.
Solution Approach 2:
The patent applies different processing qualities to different regions of the image based on depth information. Foreground regions containing meeting participants are transmitted with high quality, while background regions are excluded or transmitted with reduced quality. This local differentiation maintains essential information while eliminating harmful extraneous data.
2Adaptability or versatility
If users can select specific areas to exclude from sharing, then privacy control is improved, but system complexity increases due to depth estimation and region selection mechanisms
Solution Approach 1:
The system performs automatic monocular depth estimation and generates depth maps without requiring manual user input. The automated processing of depth information and region selection reduces the operational burden on users while maintaining high privacy control capability, thus improving adaptability without proportionally increasing complexity.
Solution Approach 2:
The patent transforms the two-dimensional image data into three-dimensional depth-space representation using monocular depth estimation. This parameter transformation enables intuitive region selection in 3D space while leveraging automated algorithms to manage the computational complexity, balancing enhanced privacy control with acceptable system complexity.
3Measurement precision
If monocular depth estimation and computer vision techniques are implemented, then accuracy in distinguishing regions of interest is improved, but computational requirements and processing time increase
Solution Approach 1:
The system performs monocular depth estimation and generates depth maps as a preliminary step before image transmission. By pre-processing the depth information and identifying regions of interest in advance, the system achieves high measurement precision for region differentiation while optimizing subsequent processing efficiency, reducing overall processing time despite the added computational step.
Data Source
Figure 1A~1B
Figure 1C~1D
Figure 2
AI summary
A system for preventing private image data captured at an endpoint from being shared during a videoconference is provided. A user can select three-dimensional regions which will not be seen during a videoconference while areas in front of the designated regions remain viewable.