Cross-Correlation Image Alignment for Multi-Modal Spectral Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image alignment technologies, such as those based on SIFT feature points, are not suitable for multi-modal multi-spectral images due to large color and gradient direction contrasts, leading to inaccurate alignment.
Innovation Solution
An image alignment method using a cross-correlation measurement model that considers both color and gradient cross-correlations between images, calculating coordinate offsets and transformations to align pixels accurately across different modalities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If SIFT feature point-based alignment technology is used, then alignment can be performed on conventional single-modal images, but alignment accuracy deteriorates when applied to multi-modal multi-spectral images with large color and gradient direction contrasts
Solution Approach 1:
The patent changes the measurement parameters from gradient-based SIFT features to cross-correlation measurements that are insensitive to gradient direction variations. By using pixel intensity cross-correlation instead of gradient magnitude and direction, the method adapts to multi-modal images while maintaining alignment accuracy despite large color and gradient contrasts between different spectral modalities.
2Measurement precision
If gradient-based feature matching is used, then alignment works well for images with similar gradient structures, but performance deteriorates when large gradient direction contrasts exist between images
Solution Approach 1:
The patent extracts and removes the gradient direction component from the matching process, retaining only the pixel intensity information for cross-correlation measurement. This extraction of the problematic gradient direction parameter allows the method to work on images with vastly different gradient directions while maintaining accuracy on images with similar gradients.
3Ease of operation
If SIFT feature point vectors are matched using Euclidean distance, then correspondence can be established between images, but matching accuracy deteriorates due to sensitivity to gradient value and gradient direction variations
Solution Approach 1:
The patent substitutes the mechanical feature extraction and matching process (SIFT keypoint detection, descriptor computation, and Euclidean distance matching) with a direct cross-correlation measurement approach. This replacement eliminates the sensitivity to gradient variations while maintaining computational simplicity, achieving both ease of operation and high matching accuracy.
Data Source
AI summary
An image alignment method and apparatus, where the method and apparatus include obtaining image information of two to-be-aligned images, determining, using a cross-correlation measurement model, first coordinate offset according to the image information of the two images, where the first coordinate offset are used to indicate position deviations of to-be-aligned pixels between the two images in the coordinate system, and aligning the two images according to coordinates of pixels in the first image in the coordinate system and the first coordinate offset.