Video Stabilization Using Zone-Of-Interest Feature Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video stabilization techniques, such as those using SIFT and SURF features, are computationally expensive and not suited for real-time applications, especially in low-cost and small video cameras, and face challenges with motion blur, noise, and incorrect feature matching due to moving objects and inliers concentrated in specific areas.
Innovation Solution
A feature-based 2-D video stabilization system that performs motion-model estimation using matched features passing a Zone-Of-Interest (ZOI) test, which ensures features are distributed across multiple zones, and employs a panning filter to distinguish between intended and unintended motion, thereby stabilizing video sequences efficiently with low power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If SIFT features are used for video stabilization, then stabilization performance is improved, but computational load increases
Solution Approach 1:
The patent replaces expensive SIFT feature detection with cheaper, simpler feature detection methods that are computationally less demanding. The invention uses alternative feature extraction techniques that provide sufficient stabilization performance without the high computational cost of SIFT, making the system suitable for real-time and low-power applications.
Solution Approach 2:
The patent modifies the feature detection parameters and methods by using zone-of-interest testing and selective feature matching strategies. This changes the approach from comprehensive SIFT feature extraction to a more targeted, efficient feature selection process that reduces computational load while maintaining stabilization effectiveness.
2Device complexity
If SURF features are used for video stabilization, then computational burden is reduced compared to SIFT, but results may still be too expensive for real-time applications
Solution Approach 1:
The patent divides the image processing into zones of interest, where feature detection and matching are performed selectively in specific regions rather than across the entire image. This segmentation approach further reduces computational burden and enables real-time processing by focusing resources on critical areas.
Solution Approach 2:
The patent performs feature matching only in zones of interest rather than across the entire image, applying partial action to achieve real-time performance. This selective approach processes only the necessary portions of the image data, reducing computational requirements while maintaining stabilization quality.
3Reliability
If feature matching is performed without zone-of-interest testing, then all features are considered, but matched features may be concentrated in specific areas leading to incorrect motion estimation
Solution Approach 1:
The patent applies different processing quality and attention to different regions of the image by implementing zone-of-interest testing. Certain zones are designated as more important for motion estimation, and feature matching is prioritized or restricted to these zones, improving reliability while managing processing complexity.
Solution Approach 2:
The patent performs preliminary zone-of-interest testing before full feature matching to identify which regions contain reliable features for motion estimation. This preliminary action filters out zones with concentrated or unreliable features, ensuring that subsequent matching operations are performed only on suitable regions.
Data Source
AI summary
According to an embodiment, a sequence of video frames as produced in a video-capture apparatus such as a video camera is stabilized against hand shaking or vibration by:—subjecting a pair of frames in the sequence to feature extraction and matching to produce a set of matched features;—subjecting the set of matched features to an outlier removal step; and—generating stabilized frames via motion-model estimation based on features resulting from outlier removal. Motion-model estimation is performed based on matched features having passed a zone-of-interest test confirmative that the matched features passing the test are distributed over a plurality of zones across the frames.