Adaptive Bounding Box Stabilization via Noise-Dependent Temporal Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video stream systems generate unstable bounding boxes due to pixel variations between frames, causing a flickering effect that degrades user experience.
Innovation Solution
Adapt the number of preceding image frames for temporal filtering of bounding boxes based on measured noise levels, increasing frames for higher noise and decreasing frames for lower noise to stabilize the bounding boxes while maintaining responsiveness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Stability of the object's composition
If temporal filtering is applied over a fixed number of preceding image frames, then bounding box stability is improved, but latency increases and responsiveness to actual object movement decreases
Solution Approach 1:
The patent applies dynamics by making the temporal filtering window adaptive rather than fixed. The number of preceding frames used for filtering is dynamically adjusted based on detected noise levels in the video stream. When noise is high, more frames are averaged to improve stability; when noise is low, fewer frames are used to reduce latency and improve responsiveness to actual movement.
Solution Approach 2:
The patent changes the parameter of temporal filtering window size based on noise level measurements. By monitoring video quality and noise characteristics, the system adjusts the filtering strength parameter to optimize the trade-off between stability and responsiveness, preventing both excessive latency and flickering effects.
2Stability of the object's composition
If temporal filtering is applied over more preceding image frames, then bounding box stability is improved in noisy conditions, but responsiveness to actual object movement decreases
Solution Approach 1:
The system dynamically adjusts the filtering window size based on real-time noise level detection. When the video stream exhibits high noise levels, a larger temporal window is used to stabilize bounding boxes. When noise levels are low, the window size is reduced to maintain responsiveness to actual object movement, thus adaptively balancing stability and speed.
Solution Approach 2:
The patent modifies the temporal filtering parameter (number of preceding frames) based on measured noise characteristics. This parameter change allows the system to optimize performance for different video quality conditions, using stronger filtering only when necessary to combat noise-induced instability.
3Speed
If no temporal filtering is applied, then responsiveness to object movement is maintained, but flickering effect occurs due to pixel variations between frames
Solution Approach 1:
The patent introduces adaptive parameter adjustment by monitoring noise levels and modifying the temporal filtering strength accordingly. This allows the system to apply filtering only when noise-induced flickering is detected, maintaining responsiveness during clean video segments while stabilizing bounding boxes during noisy segments.
Solution Approach 2:
The system uses feedback from noise level measurements to control the temporal filtering application. By continuously monitoring video quality and adjusting filtering strength based on this feedback, the system eliminates flickering effects caused by pixel variations while preserving responsiveness to actual object movement.
Data Source
Figure 1~2A
Figure 2B~2C
Figure 3~4
AI summary
The present disclosure relates to a method of stabilizing bounding boxes for objects in a video stream, the method comprising: receiving a video stream comprising a sequence of image frames; detecting an object in the image frames and generating a bounding box surrounding the object; measuring a noise level for the video stream; and temporally filtering the bounding box over a plurality of image frames based on the measured noise level, thereby stabilizing the bounding box in the video stream. The disclosure further relates to an image processing system.