Orthogonal Disparity Decomposition for Stereo Depth Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computational stereo methods face challenges in real-time processing and accurately constructing depth maps from stereo images, especially in noisy and low-color-fidelity environments, as they rely heavily on global knowledge and fail to effectively combine local and global constraints.
Innovation Solution
The approach involves orthogonal and collaborative disparity decomposition, which segments images into orthogonal spaces and uses alpha-encoded segment flags to ensure orthogonality, allowing for efficient and accurate depth map updates by focusing on significant changes in stereo image sequences and exploiting scene redundancy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If global knowledge propagation is used to construct disparity maps, then depth map coverage is improved, but computational complexity increases and real-time processing becomes difficult
Solution Approach 1:
The patent segments the disparity computation process into multiple independent components: stable segments that remain unchanged across frames, unstable segments that require updates, and background segments. This segmentation allows the system to process only relevant portions of the image at full resolution while using lower resolution for stable regions, significantly reducing computational complexity while maintaining depth map accuracy.
Solution Approach 2:
The patent performs preliminary segmentation and identification of stable/unstable regions before conducting full disparity computation. By pre-identifying which segments require processing and preparing alpha-encoded segment flags in advance, the system avoids redundant computations and achieves real-time processing capability while maintaining accurate depth maps.
2Measurement precision
If color-based segmentation is used to improve disparity computation, then local constraint accuracy is improved, but reliability decreases in noisy and low-color-fidelity environments
Solution Approach 1:
The patent changes the parameter space for segmentation by using alpha-encoded segment flags and stability-based classification instead of relying solely on color information. This allows the system to maintain segmentation reliability in noisy environments by using multiple parameters including temporal stability, spatial coherence, and confidence metrics rather than depending on color fidelity alone.
Solution Approach 2:
The patent introduces an intermediary stability assessment mechanism that mediates between color-based segmentation and final disparity computation. The alpha-encoded segment flags act as an intermediary layer that validates and refines segment boundaries, ensuring reliable segmentation even when color information is noisy or low fidelity.
3Measurement precision
If full image processing is performed for each frame, then depth map accuracy is maintained, but processing speed decreases and real-time performance is lost
Solution Approach 1:
The patent implements dynamic processing where the level of processing applied to each segment varies based on its stability characteristics. Stable segments processed at lower resolution or with simplified algorithms, while unstable segments receive full processing. This dynamic adaptation maintains depth map accuracy for critical regions while achieving real-time processing speeds through selective processing.
Solution Approach 2:
The patent applies partial processing to stable segments that don't require full analysis, and excessive (full) processing only to unstable segments that need detailed examination. This selective application of processing intensity maintains accuracy where needed while dramatically improving overall processing speed and enabling real-time performance.
Data Source
AI summary
A novel disparity computation technique is presented which comprises multiple orthogonal disparity maps, generated from approximately orthogonal decomposition feature spaces, collaboratively generating a composite disparity map. Using an approximately orthogonal feature set extracted from such feature spaces produces an approximately orthogonal set of disparity maps that can be composited together to produce a final disparity map. Various methods for dimensioning scenes and are presented. One approach extracts the top and bottom vertices of a cuboid, along with the set of lines, whose intersections define such points. Tt then defines a unique box from these two intersections as well as the associated lines. Orthographic projection is then attempted, to recenter the box perspective. This is followed by the extraction of the three-dimensional information that is associated with the box, and finally, the dimensions of the box are computed. The same concepts can apply to hallways, rooms, and any other object.