Image Registration Using Dual Transform Models for Parallax Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image registration methods in technical applications like surveillance and medical imaging face challenges in accurately registering images from sensors at different distances without parallax errors, which affects the precision and reliability of algorithms for variation detection, motion tracking, and object recognition.
Innovation Solution
The proposed solution involves an apparatus and method that uses a first transform model estimator and a second transform model estimator to generate respective transform models based on feature points between images, with a registrator transforming partial images to register thermal and visible light images without parallax errors, employing algorithms like SIFT and RANSAC for feature extraction and model estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single transform model is used for image registration, then the device complexity is low, but registration precision deteriorates due to parallax errors between objects at different distances
Solution Approach 1:
The patent divides the image registration process into multiple segments: a first transform model for global alignment and a second transform model for local refinement. This segmentation allows different transformation strategies to be applied to different regions, improving registration precision while managing complexity through modular processing.
Solution Approach 2:
The patent applies local quality by using a second transform model specifically for regions containing objects at different distances. This local refinement approach针对性地 addresses parallax errors in critical regions without requiring complete re-registration of the entire image, thus improving precision where needed while controlling overall complexity.
2Measurement precision
If multiple transform models are used to account for varying object distances, then registration precision improves, but device complexity increases
Solution Approach 1:
The patent performs preliminary action by first applying the first transform model for global alignment before applying the second transform model for local refinement. This sequential approach prepares the images in stages, reducing the complexity of each individual transformation step while achieving high overall precision.
Solution Approach 2:
The patent applies partial action by focusing the computationally intensive second transform model only on regions containing objects at different distances, rather than processing the entire image. This selective approach improves registration precision for critical regions while minimizing the overall computational complexity.
3Measurement precision
If partial image segmentation is performed to generate transform models, then registration accuracy improves, but processing time increases
Solution Approach 1:
The patent applies partial action by performing detailed transform model estimation only on segmented regions containing objects at different distances, rather than processing the entire image. This reduces processing time while maintaining high registration accuracy for critical regions.
Solution Approach 2:
The patent performs preliminary segmentation and identification of regions requiring detailed registration before applying computationally intensive transform models. This preliminary action prepares the data in advance, allowing faster processing during the actual registration phase while maintaining high accuracy.
Data Source
AI summary
There are provided an apparatus and method for registering images. The apparatus includes at least one processor configured to implement: a first transform model estimator configured to generate a first transform model based on corresponding feature points between a first image and a second image; a second transform model estimator configured to generate a second transform model based on corresponding feature points between a first partial image of the first image and a second partial image of the second image, the second partial image being generated based on the first partial image; and a registrator configured to register the first image and the second image by transforming the first partial image using the first transform model and the second transform model.