Video Frame Deflection Filtering for Stable Pose-Based Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video frames in video processing are affected by noise, leading to video distortion and instability due to imbalanced pose estimation errors across frames.
Innovation Solution
A method and apparatus for video frame processing that involves extracting poses of a key part, determining deflection degree, generating a control signal, performing backlash filtering to obtain updated deflection values, and performing image processing based on these values to stabilize the video frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If video frames are processed directly without filtering, then processing speed is maintained, but video distortion and instability occur due to noise-induced pose estimation errors
Solution Approach 1:
The patent applies preliminary action by performing backlash filtering on control signals before using them for video frame processing. The filtering operation removes noise-induced jitter from pose estimation data in advance, ensuring that subsequent image processing operations work with stable, filtered data rather than raw noisy data. This preliminary filtering step prevents distortion from propagating through the processing pipeline.
Solution Approach 2:
The patent introduces an intermediary element - the backlash filter - that mediates between the noisy pose estimation output and the video frame processing input. This filter acts as a buffer that transforms the unstable control signals into stable filtered signals, allowing the system to maintain reliability without requiring complete redesign of the processing architecture.
2Reliability
If backlash filtering is applied to control signals, then video frame stability is improved, but processing time increases
Solution Approach 1:
The patent applies parameter changes by adjusting the backlash filtering coefficient to optimize the balance between filtering effectiveness and processing speed. By tuning this parameter, the system can achieve adequate noise reduction with minimal processing overhead, allowing real-time or near-real-time video stabilization without excessive computational delay.
3Measurement precision
If pose estimation is performed on all video frames, then accuracy is improved, but computational load increases due to noise in each frame
Solution Approach 1:
The patent applies the extraction principle by separating the pose estimation function from the video processing function. Instead of performing full pose estimation on every noisy video frame, the system extracts key pose information and applies backlash filtering to the extracted control signals. This extraction approach reduces computational energy by avoiding redundant processing of noisy data while maintaining the essential pose information needed for stabilization.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
The present application relates to a video frame processing method and apparatus, a computer device, a storage medium, and a computer program product. The method can be applied to scenarios such as cloud conferencing, cloud storage, artificial intelligence, and intelligent transportation. The method comprises: receiving a plurality of video frames carrying noise, the plurality of video frames comprising an object (step S202); extracting postures of a key part of the object in the plurality of video frames (step S204); according to the postures of the key part, determining deflection degree values of the key part in the plurality of video frames (step S206); generating a control signal on the basis of the deflection degree values (step S208); performing background filtering on the control signal to obtain a deflection update value used for representing processing of the plurality of video frames (step S210); and sequentially performing image processing on the plurality of video frames on the basis of the deflection update value (step S212).