Video Stream Cropping for Equal-Size Conference Participant Views
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video conferencing systems struggle to provide equal-sized views of multiple participants, especially when there are varying numbers of people at different locations, leading to inefficiencies and complex installations or incomplete region-of-interest focus.
Innovation Solution
A method and system that detects predefined objects in a video stream, selects crop regions around these objects, and transmits them as separate or composite views, ensuring equal-sized representations of participants regardless of their number or distance from the camera.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If multiple video cameras are deployed at a location with multiple people to capture separate video streams, then each person can be shown at a similar size, but the system complexity and installation cost increase significantly
Solution Approach 1:
The patent divides a single video stream into multiple cropped regions, each focusing on a different person or object of interest. The processor identifies multiple objects in the video stream and generates separate cropped video streams for each object, allowing each participant to be displayed at a consistent size without requiring multiple physical cameras.
Solution Approach 2:
The patent extracts specific regions of interest from the full video stream by cropping around identified objects. This extraction process isolates each person or object of interest into its own video stream, enabling focused display of multiple participants at equal sizes while using only a single camera.
2Device complexity
If a single camera is used at a location with multiple people, then the installation remains simple, but all people appear much smaller in the video stream compared to locations with single users
Solution Approach 1:
The patent segments the single video stream into multiple focused cropped regions, each highlighting a specific person or object. This segmentation allows the system to maintain installation simplicity while improving video quality by ensuring each participant appears at a consistent, prominent size rather than all appearing small in a wide shot.
Solution Approach 2:
The patent transforms the spatial arrangement problem by creating multiple virtual video streams from a single physical camera. Instead of adding more cameras in the physical dimension, the system operates in the digital signal processing dimension, generating multiple cropped views that simulate the effect of multiple cameras.
3Ease of operation
If speaker tracking is used to focus on the speaking participant, then the region of interest is emphasized, but other areas that could be regions of interest are removed from the full picture
Solution Approach 1:
The patent creates a multi-functional system where a single video stream serves multiple purposes simultaneously. It generates multiple cropped regions that can each focus on different objects or speakers, allowing the system to adapt to different situations (different speakers, different objects of interest) without losing information about other potential regions of interest.
Solution Approach 2:
The patent implements a dynamic system where the cropped regions can change based on which objects are currently speaking or are of interest. The system can dynamically adjust which regions are generated and transmitted, allowing flexible focus on speakers while maintaining the option to show other regions when appropriate.
Data Source
AI summary
A method for manipulating an initial video stream captured by a camera in a video conferencing endpoint into multiple views corresponding to regions of interest. The method comprises detecting objects of one or more predefined types in a frame of the initial video stream; selecting a plurality of crop regions (2200a, 2200b, 2200c) from the frame of the initial video stream, and transmitting the plurality of crop regions. Each crop region includes at least one bounding box and each bounding box includes a detected object of a predefined type.