Multiview Video Stabilization Using Disparity Map Propagation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques are inadequate for performing high-quality and efficient multiview video stabilization, particularly in immersive head-mounted display and lightfield display environments, where multiview video captured by handled devices often appears jittery and unsuitable for user enjoyment.
Innovation Solution
The proposed solution involves performing single view video stabilization on a reference video stream to generate a stabilized reference stream, which is then propagated to other video streams by generating a target disparity map and minimizing vertical or horizontal components, ensuring consistent stabilization across multiple video streams, thereby reducing eye fatigue and improving video quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiview video is captured by handled devices to provide immersive experiences, then viewing creativity and immersion are improved, but video stability deteriorates causing jittery footage
Solution Approach 1:
The patent segments the multiview video stabilization problem into independent per-view stabilization tasks. Each video stream is stabilized separately using its own motion estimation and compensation processes, allowing individual optimization without affecting other views. This segmentation enables handling of complex multiview scenarios while maintaining overall system stability.
Solution Approach 2:
The patent introduces disparity maps as intermediary structures that capture depth information between different video streams. These disparity maps serve as mediators to guide the stabilization process, enabling the system to understand spatial relationships and apply appropriate stabilization transformations that maintain multi-view geometry consistency across all views.
2Stability of the object's composition
If traditional stabilization techniques are applied to multiview video, then some stabilization effect is achieved, but quality and consistency across multiple views deteriorate
Solution Approach 1:
The patent applies local quality by allowing different stabilization parameters and transformations for different video streams based on their specific characteristics. Each view can have customized motion compensation tailored to its unique motion patterns and depth relationships, ensuring high stabilization quality for each individual stream while maintaining overall multiview consistency.
Solution Approach 2:
The patent dynamically adjusts stabilization parameters including motion vectors, disparity map resolutions, and transformation matrices based on scene content and depth information. These parameter changes enable adaptive stabilization that maintains high quality across varying video conditions while preserving the three-dimensional structure of multiview content.
3Ease of manufacture
If stabilization is processed for each video stream independently, then processing simplicity is improved, but computational efficiency deteriorates
Solution Approach 1:
The patent merges common processing operations across multiple video streams, such as disparity map generation and motion estimation algorithms. By sharing computational resources and intermediate results between views, the system achieves the simplicity of independent processing while gaining the efficiency of coordinated computation, reducing redundant calculations across streams.
Data Source
AI summary
Techniques related to multiview video stabilization are discussed. Such techniques may include performing single view video stabilization on a reference video stream and propagating the single view video stabilization to another video stream based on a target disparity map.