Adaptive Disparity Range Detection for Stereo Content
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing stereo video processing methods require a large amount of computation for disparity map estimation, often leading to inefficiencies such as getting stuck in local minimums, missing actual disparities, and performing unnecessary computations due to a constant disparity search range, which is not adaptive to different scenes.
Innovation Solution
A method and device that detect the disparity range by calibrating stereo content, identifying distinctive pixels, calculating cost values, determining reliable disparity values, and creating a histogram to determine the minimum and maximum disparity candidates, thereby reducing computational complexity and improving 3D imaging.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a constant predetermined disparity search range is used for all scenes, then the disparity search can be performed uniformly, but unnecessary computations are performed for out of range disparity candidates and the algorithm stacks at local minimums
Solution Approach 1:
The patent applies dynamics by making the disparity search range adaptive rather than constant. The system dynamically adjusts the search range based on scene characteristics by detecting distinctive pixels and analyzing their disparity distributions. This allows the search range to change according to the specific scene content, avoiding unnecessary computations in regions where disparities are known to be small or absent.
Solution Approach 2:
The patent implements local quality by performing disparity search with different ranges for different regions of the image. Instead of applying a uniform search range to the entire image, the system identifies regions with distinctive pixels and applies localized search strategies. This allows computationally intensive searches only where needed while using smaller or no search ranges in regions where disparities are unlikely or already known.
2Reliability
If a large disparity search range is used to cover all possible disparities, then actual disparities are less likely to be missed, but the computation time increases significantly
Solution Approach 1:
The patent applies preliminary action by performing disparity range detection before the full disparity map estimation. The system first identifies distinctive pixels and determines their disparity ranges, then uses this information to guide the subsequent disparity search. This preliminary analysis allows the system to focus computational resources on relevant disparity ranges, avoiding unnecessary searches in regions where disparities are constrained.
Solution Approach 2:
The patent implements partial action by performing disparity search only for regions containing distinctive pixels rather than for all pixels in the image. The system identifies and processes only the subset of pixels that provide useful disparity information, leaving other regions unprocessed or processed with reduced effort. This selective approach maintains reliability for important features while significantly reducing overall computation time.
3Measurement precision
If disparity search is performed independently for each pixel, then precise disparity can be obtained, but the computational complexity becomes extremely high
Solution Approach 1:
The patent applies merging by combining multiple pixels into processing units or blocks. Instead of treating each pixel independently, the system groups pixels spatially and processes them together, sharing computational resources and intermediate results. This approach maintains precision by considering local pixel relationships while reducing overall complexity through shared computations across pixel groups.
Solution Approach 2:
The patent implements segmentation by dividing the image into regions based on distinctive pixel locations and disparity characteristics. The system processes each segment with appropriate search strategies rather than applying a uniform approach to the entire image. This segmentation allows precise disparity estimation in critical regions while using coarser or no search in less important regions, balancing precision and complexity.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
Present invention provides a method and a device to detect a disparity range of stereo content. Said method comprises the steps of; calibrating the stereo content (10), detecting distinctive pixels in the stereo content (30), calculating cost values (40), determining the candidate disparity values (50), and creating a histogram (60). Said device comprises; at least one stereo calibrator (1), at least one distinctive pixel detector (3), at least one cost calculator (4), means for determining the candidate disparity values (5), means for creating a histogram (6), and at least one histogram thresholder (7).