Stereo Correspondence Search Using Multi-Level Aggregate Digital Data Structure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for finding stereo correspondence between images are inefficient and costly in terms of computational resources, particularly when dealing with multiple resolutions and variance awareness, often leading to missed high-frequency details and incorrect error minimization.
Innovation Solution
The use of a multi-level aggregate digital data structure (MLADDS) that represents images at different resolutions, allowing for efficient comparison and error calculation across multiple resolutions, and variance-aware processing to optimize computational efforts by rejecting ineligible elements and focusing on high-variance areas for further refinement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional stereo correspondence search methods are used, then computational accuracy can be maintained, but computational cost and processing time increase significantly
Solution Approach 1:
The patent divides the search space into multiple regions or blocks, allowing the stereo correspondence search to be performed independently in each region. This segmentation enables parallel processing and reduces the overall computational complexity while maintaining accuracy through localized searches.
Solution Approach 2:
The patent performs preliminary processing steps such as generating candidate correspondence lists, computing disparity ranges, or creating coarse matching results before executing the full stereo correspondence search. These preliminary actions narrow down the search space and reduce the computational burden of subsequent detailed matching operations.
2Measurement precision
If high-resolution processing is applied throughout, then detail accuracy improves, but computational resources are wasted on low-variance areas
Solution Approach 1:
The patent applies different processing resolutions or levels of detail to different regions of the image based on their variance characteristics. High-variance regions (containing important details or edges) receive high-resolution processing, while low-variance regions (smooth areas) are processed at lower resolutions or skipped entirely, optimizing the balance between accuracy and computational efficiency.
Solution Approach 2:
The patent dynamically adjusts processing parameters such as block size, search range, or matching algorithm complexity based on local image variance. Regions with high variance use more computationally intensive parameters to capture details, while low-variance regions use simplified parameters to reduce computational overhead.
3Reliability
If exhaustive search is performed across all search domains, then no correspondences are missed, but processing time becomes prohibitive
Solution Approach 1:
The patent performs a partial search by limiting the search to a subset of the full search domain based on preliminary analysis, such as disparity constraints, occlusion masks, or region-of-interest identification. This partial action approach finds the most important correspondences without exhaustively searching every possible location, significantly reducing processing time while maintaining reliability for critical features.
Solution Approach 2:
The patent uses preliminary analysis steps to identify promising search regions, constrain disparity ranges, or eliminate impossible correspondences before executing the main search algorithm. These preliminary actions reduce the effective search space and allow the system to achieve reliable results faster by focusing computational effort on the most likely correspondence locations.
Data Source
AI summary
Methods, systems, devices and computer software/program code products enable efficiently finding stereo correspondence between a feature or set of features in a first image or signal, and a search domain in a second image or signal.