Video Key Point Stabilization Using Weighted Smoothing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video image processing technologies face challenges with slow processing speed and poor effectiveness in preventing video jitter, particularly in human face and body posture key point detection, leading to a poor user experience.
Innovation Solution
A video image processing method that uses weighted smoothing of key points based on historical key point queues and weight queues, where weights are assigned according to the timing sequence, with closer frames having higher weights and farther frames having lower weights, to adjust and stabilize key points in video frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If Kalman filtering or optical flow method is used to predict video frames, then video jitter can be prevented, but processing speed becomes slow and time overhead increases
Solution Approach 1:
The patent extracts only the essential key point coordinate information from video frames and performs smoothing directly on these coordinates, rather than using comprehensive filtering methods like Kalman filtering that process entire frames or complex optical flow calculations. This extraction of core information enables faster processing while maintaining jitter prevention effectiveness.
Solution Approach 2:
The patent uses a simple weighted average calculation with a limited historical window (N frames) that can be quickly computed and discarded, replacing the need for complex, computationally expensive filtering algorithms. The lightweight calculation approach processes data rapidly with minimal time overhead.
2Reliability
If Kalman filtering or optical flow method is used to predict video frames, then video jitter can be prevented, but time overhead becomes large
Solution Approach 1:
The patent pre-calculates and stores key point coordinates from historical frames in a buffer, making this data readily available for rapid weighted average calculations. By preparing the historical data in advance and organizing it efficiently, the system minimizes computation time during actual jitter prevention processing.
Solution Approach 2:
The patent employs a simple weighted average formula that requires minimal computational resources and can be executed quickly, replacing time-consuming filtering algorithms. The calculation uses a fixed window of historical data that is processed and discarded efficiently, reducing overall time overhead.
3Productivity
If simple key point smoothing is applied, then processing speed is fast, but video jitter prevention effectiveness is poor
Solution Approach 1:
The patent dynamically adjusts the weight assigned to historical key point coordinates based on their temporal distance from the current frame. Closer frames receive higher weights while older frames receive lower weights, creating an adaptive smoothing effect that responds to changing video content and maintains effectiveness without sacrificing processing speed.
Solution Approach 2:
The patent changes the parameter of weight distribution across historical frames, using a weighted average where weights decrease with temporal distance. This parameter adjustment allows the system to emphasize recent frames while still incorporating historical data, achieving both fast processing and effective jitter prevention.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Embodiments of this application disclose a video image processing method and apparatus, and relate to the image processing field. The method in the embodiments of this application includes: obtaining a first key point of an Mth frame image in a video; performing weighted smoothing on the first key point of the Mth frame image according to first key points in a historical key point queue and weights in a first target weight queue, to obtain a target key point, the historical key point queue including a first key point corresponding to each frame image in N frame images, the N frame images being images before the Mth frame image, N > 0, the weights in the first target weight queue corresponding to the first key points in the historical key point queue, a weight corresponding to a first key point of an (M-a)th frame image being greater than or equal to a weight corresponding to a first key point of an (M-b)th frame image, and a < b ; and adjusting the Mth frame image according to the target key point. In this application, the effect in preventing video jitter is remarkable, the algorithm is simple, and the processing speed is fast.