Stereo Image Matching via Mean Field Approximation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current stereo image matching methods are inefficient in reducing computation costs, leading to errors due to occlusions and inconsistencies in binocular disparity maps, particularly when dealing with scenes where geometry consistency is compromised.
Innovation Solution
A method that calculates data and smoothness costs for pixel matching using binocular disparity information, employing a mean field approximation to maximize posterior probability distributions and minimize Kullback-Leibler divergence, with adjustable initial costs and regularization coefficients, while considering geometric relationships and neighborhood similarities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional stereo image matching methods are used, then matching results can be obtained, but computation costs are high and errors occur due to occlusions and inconsistencies
Solution Approach 1:
The patent segments the stereo matching problem into distinct cost components: data cost calculation based on geometric relationships and smoothness cost calculation based on neighborhood similarities. This segmentation allows independent optimization of each cost function and enables efficient computation through mean field approximation, resolving the contradiction between accuracy and computation cost
Solution Approach 2:
The patent transforms the stereo matching problem by changing parameters from direct pixel comparison to probability distribution optimization. By maximizing posterior probability distributions and minimizing Kullback-Leibler divergence, the method achieves higher accuracy while reducing computation cost through efficient parameter optimization
2Measurement precision
If detailed geometric relationships are considered for accurate matching, then matching precision improves, but computation complexity increases
Solution Approach 1:
The patent changes the approach from direct geometric constraint satisfaction to probability-based parameter optimization. By using posterior probability distributions and Kullback-Leibler divergence minimization, the method maintains high measurement precision while reducing algorithmic complexity through mean field approximation
Solution Approach 2:
The patent substitutes direct geometric constraint enforcement with a statistical mechanics approach using probability distributions and energy minimization. This substitution simplifies the algorithmic complexity while maintaining precision by replacing rigid geometric rules with flexible probabilistic models
Data Source
AI summary
A method for matching stereo images may calculate a data cost value for each of a plurality of images, calculate a smoothness cost value for each of the plurality of images, and match pixels among the plurality of images based on the data cost value and the smoothness cost value.