Image Correspondence Analysis Using Even-Odd Convolution Kernels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current correspondence analysis techniques in stereo imaging are inefficient, noisy, and fail to achieve high resolution and accuracy in real-time processing due to high computational complexity and suboptimal signal-to-noise ratios, particularly in handling disparities and 3D data generation.
Innovation Solution
A correspondence analyzer that combines information from both spatial and spatial frequency windows using specially defined convolution kernels to determine disparities with low noise, employing even and odd harmonic functions for precise disparity calculation and confidence assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If current correspondence analysis techniques are used to identify point correspondences in stereo images, then 3D data can be generated, but the processing is memory- and calculation-intensive, slowing down real-time performance
Solution Approach 1:
The patent divides the image processing task into smaller patches (e.g., 16x16 or 32x32 pixel regions) and processes them independently using parallel computation units. This segmentation allows the system to handle multiple patches simultaneously, reducing overall processing time and enabling real-time performance while maintaining accuracy through localized correlation analysis.
2Measurement precision
If large image patches are processed to reliably identify correspondences, then measurement accuracy improves, but memory and calculation requirements increase significantly
Solution Approach 1:
The patent employs a multi-resolution approach where coarse disparity estimation is performed first on smaller patches, followed by refined estimation on selected regions. This partial processing strategy achieves high measurement precision for critical correspondences while avoiding the excessive computational cost of processing all regions at maximum resolution, thus optimizing the trade-off between accuracy and resource consumption.
3Measurement precision
If phase information is calculated independently in multiple spatial frequency ranges to achieve high resolution, then spatial frequency resolution improves, but noise increases and signal information is lost
Solution Approach 1:
The patent combines phase information from multiple spatial frequency ranges using a weighted integration approach. Instead of treating each frequency range independently, the system merges their contributions while applying adaptive weighting to emphasize reliable frequency components and suppress noisy ones. This merging strategy preserves high spatial frequency resolution while maintaining an optimal signal-to-noise ratio through cooperative integration of multi-frequency data.
Data Source
AI summary
A device and a method that can be used to perform a correspondence analysis in image data in a particularly low-noise and efficient manner. The device and method include selecting image patches from individual images and generating a plurality of one-dimensional signals using even and odd convolution kernels in the spatial window in each case, and processing differences between the convolution results in a non-linear manner and accumulating these differences to form a correspondence function and evaluating this function.