Temporal Consistency in 3D Video Disparity Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The computation of disparities in three-dimensional video imaging is not exact and secure, leading to artifacts like flickering and false colors in video sequences due to unstable disparity estimations, especially in complex image zones with uniform colors and no precise contours.
Innovation Solution
A method that detects still zones in image pairs by comparing current and previous images, assigns modified disparity values to pixels in still zones, and uses bilateral filtering to refine the detection of invariant zones, ensuring stability and coherence across frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If disparity estimation is performed for each pair of images independently, then the computation is simple and fast, but the disparity values become unstable and produce artifacts like flickering and false colors
Solution Approach 1:
The patent applies preliminary action by detecting still zones in advance before disparity computation. The method identifies regions that remain unchanged across multiple frames and pre-mark them for special handling. This allows the disparity estimation to focus computational resources only on moving regions while maintaining stable values in still zones, thereby improving reliability without proportionally increasing overall complexity.
Solution Approach 2:
The patent implements local quality by applying different disparity computation strategies to different regions of the image. In still zones, the method maintains previous disparity values or applies constrained estimation, while in moving zones it performs full disparity estimation. This region-specific approach ensures stable disparity values where needed without unnecessarily complicating the overall computation process.
2Measurement precision
If disparity estimation is performed using correlation of image portions, then the method can handle complex image zones, but the estimation becomes imprecise in zones with uniform colors and no precise contours
Solution Approach 1:
The patent applies feedback by using temporal information from previous frames to guide current disparity estimation. The method detects still zones by comparing current frame with previous frames, then uses this feedback to constrain disparity estimation in those zones. This feedback mechanism provides additional constraints that help resolve ambiguities in uniform regions, improving measurement precision without increasing the difficulty of detection.
3Reliability
If the computer assigns disparity values for every pixel in each frame, then the disparity map is complete, but variations in estimations between frames cause flickering and discomforting artifacts
Solution Approach 1:
The patent applies segmentation by dividing the image into still zones and moving zones based on temporal comparison. This segmentation allows the system to apply different processing strategies: maintaining stable disparity values in still zones and performing full estimation in moving zones. The result is improved temporal consistency without the computational cost of processing every pixel equally in every frame.
Solution Approach 2:
The patent implements partial action by performing full disparity estimation only where necessary (in moving zones) and using simplified or maintained values in still zones. This selective approach achieves temporal consistency and reduces artifacts without the computational expense of exhaustive processing, thereby maintaining productivity while improving reliability.
Data Source
AI summary
The invention relates to three dimensional video imaging, in which at least a left view and a right view of a moving scene are produced and a map of disparities is produced for all pixels of the successive images of a video sequence.