Scene Segmentation Map Updates Triggered by Camera Focus Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image segmentation, particularly scene segmentation, is a processing-demanding task that poses limitations when performed within a camera device, especially during live video streaming, necessitating more efficient methods for updating scene segmentation maps.
Innovation Solution
A method and device that utilize focus value changes to trigger scene segmentation map updates, aligning the map to new scene characteristics by performing image analysis only when significant focus deviations occur, optimizing processing efficiency by dividing the image analysis into object segmentation and depth estimation parts and executing them sequentially.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If image segmentation is performed continuously to keep the scene segmentation map updated, then the scene segmentation map remains current and accurate, but the processing demand and resource consumption increase substantially
Solution Approach 1:
The patent implements periodic action by triggering image segmentation only at specific intervals - when focus changes are detected - rather than continuously. The system monitors focus values and initiates segmentation processing only when the focus changes exceed a predetermined threshold, creating a event-driven periodic update mechanism that reduces unnecessary processing while maintaining map accuracy.
Solution Approach 2:
The patent applies parameter changes by using focus value changes as a trigger condition for segmentation updates. Instead of using a fixed time-based update schedule, the system dynamically adjusts processing frequency based on the parameter (focus value) changes in the scene, updating the segmentation map only when focus deviations exceed a threshold, thereby adapting processing intensity to actual scene changes.
2Speed
If image segmentation is performed in real-time during live video streaming, then the scene segmentation map is updated promptly, but the processing demand exceeds the capabilities of edge devices
Solution Approach 1:
The patent implements partial action by performing image segmentation only on portions of the image when focus changes are detected, rather than processing the entire image continuously. The system selectively triggers segmentation based on focus threshold deviations, performing partial updates only when and where needed, reducing the overall processing burden on edge devices while maintaining timely updates for changed regions.
3Reliability
If the scene segmentation map is updated frequently, then the map reflects current scene characteristics accurately, but unnecessary processing occurs when the scene has not changed
Solution Approach 1:
The patent implements feedback by continuously monitoring focus values and using this information to control segmentation updates. The system establishes a feedback loop where focus change detection triggers segmentation only when actual scene changes occur, preventing unnecessary processing during stable periods. This feedback mechanism ensures the segmentation map is updated in response to real scene changes rather than on a fixed schedule.
Data Source
Figure 1~2
Figure 3
Figure 4a~5
AI summary
The application describes a method (300) and processing device (46) for providing a scene segmentation map for a total field of view (12) of an image capturing device (10). The method (300) comprises obtaining (301) a scene segmentation map for the total field of view (12); receiving (302) an indication that a new focus value has been set for the image capturing device (10) for acquiring images of a current field of view (14); comparing (303) the new focus value with a stored focus value associated with a focus region (22) in the current field of view (14), wherein the stored focus value represents a previously used focus value; upon the new focus value deviating from the stored focus value more than a trigger threshold, triggering (305) a scene segmentation map update process; and otherwise maintaining (309) the scene segmentation map.