Selective Background Object Segmentation in Video Conferencing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video conferencing systems either require universal blurring or replacement of the entire background environment, which may not allow users to selectively conceal personal items and areas without altering the overall aesthetic or presenting an inaccurate background.
Innovation Solution
An image alteration system that segments the background environment to identify and modify specific objects, allowing users to blur, replace, or omit personal items and areas without changing the appearance of other objects, using a processor-controlled system with memory storage for image processing and user input options.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If universal blurring or replacement of the entire background environment is applied, then personal information concealment is improved, but background environment authenticity and aesthetics deteriorate
Solution Approach 1:
The system segments the background environment into multiple detectable objects and regions, allowing selective application of blurring or replacement only to specific objects containing personal information, rather than applying universal processing to the entire background. This enables targeted concealment while preserving the authenticity and aesthetics of the overall background environment.
Solution Approach 2:
The system applies different processing qualities to different regions of the background environment - applying blurring or replacement only to local regions containing personal information while leaving other regions unchanged. This creates local quality variation that conceals sensitive information while maintaining the overall authenticity and visual appeal of the background.
2Adaptability or versatility
If selective object modification is implemented, then background environment authenticity is improved, but system complexity increases
Solution Approach 1:
The system employs automated object detection and classification algorithms that automatically identify and categorize objects in the background environment without requiring manual intervention. The system self-services by autonomously determining which objects contain personal information and applying appropriate processing, thereby reducing operational complexity despite the sophisticated processing required.
Solution Approach 2:
The system utilizes adjustable processing parameters such as blur intensity, replacement confidence thresholds, and object detection sensitivity that can be modified to balance processing effectiveness with computational complexity. By changing these parameters, the system can adapt to different complexity requirements while maintaining selective object modification capabilities.
3Measurement precision
If high-resolution camera feeds are used, then image quality is improved, but personal information disclosure risk increases
Solution Approach 1:
The system introduces an intermediary processing layer between the high-resolution camera feed and the transmitted video stream. This intermediary layer detects and processes objects containing personal information by applying blurring or replacement, thereby mediating between the high image quality requirement and the personal information protection need. The high-resolution feed is processed selectively to remove harmful disclosure risks while preserving overall image quality.
Data Source
AI summary
An image alteration system includes a memory and one or more processors. The processors receive input image data that depicts a background environment, and segment the input image data to locate objects depicted in the background environment. The processors identify a set of one or more of the objects to modify, and generate output image data that depicts a modified version of the background environment for display on a display device. The processors change an appearance of the one or more objects in the set as depicted in the output image data, relative to the appearance of the one or more objects as depicted in the input image data, without changing an appearance of one of more other objects located in the background environment. The input image data and the output image data represent video streams.