Real-Time Large-Field-Angle Image Stitching via Segmented Graph Cut
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image stitching methods for wide-angle camera systems face limitations in real-time performance due to high computational complexity and unnatural seam visibility, especially when dealing with large disparities between images.
Innovation Solution
A method that pre-calculates the homography matrix using camera parameters and improves the graph cut algorithm to achieve real-time stitching, involving feathering fusion and histogram equalization to create a natural stitching result, by dividing images into blocks and using a graph cut algorithm to find an optimal stitching seam.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional graph cut algorithm is used for optimal stitching seam, then stitching quality is improved, but real-time performance deteriorates
Solution Approach 1:
The patent divides the image into multiple blocks and performs graph cut algorithm on each block separately rather than the entire image at once. This segmentation reduces the computational complexity of the graph cut algorithm while maintaining stitching quality, enabling real-time performance.
Solution Approach 2:
The patent pre-calculates the homography matrix using camera parameters before performing the graph cut algorithm. This preliminary action provides a good initial alignment that reduces the search space for the graph cut algorithm, significantly improving processing speed while maintaining stitching quality.
2Measurement precision
If feature point matching is performed for image registration, then stitching accuracy is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary camera calibration to obtain internal parameters (focal length, optical center) and external parameters (rotation, translation) before image stitching. The homography matrix is then calculated directly from these calibrated parameters, eliminating the need for time-consuming feature point matching while maintaining high stitching accuracy.
3Area of stationary object
If images are stitched with large disparity, then field angle coverage is improved, but stitching complexity increases
Solution Approach 1:
The patent calculates the homography matrix in advance using camera calibration parameters, which provides a robust framework for handling large disparities. This preliminary calculation simplifies the subsequent stitching process by establishing accurate geometric relationships between images with large angular differences.
Solution Approach 2:
The patent divides images into blocks and processes them separately with graph cut algorithm, which reduces the complexity of handling large-disparity stitching while maintaining quality across the entire field of view.
4Manufacturing precision
If mesh optimization method is used for stitching, then stitching effect is improved, but calculation complexity increases
Solution Approach 1:
The patent divides the image into C1*C2 mesh blocks and performs graph cut algorithm on each block separately. This segmentation reduces the overall calculation complexity compared to applying mesh optimization to the entire image, while still achieving good stitching effects through local optimization.
Solution Approach 2:
The patent pre-calculates the homography matrix to provide a good initial alignment, which reduces the iteration次数 and computational burden of the subsequent mesh optimization process, achieving a balance between stitching effect and calculation complexity.
Data Source
AI summary
The present invention discloses a large-field-angle image real-time stitching method based on calibration. First, a calibration algorithm is used to solve the positional relationship between cameras, and the prior information is used to solve a homography matrix between images. The system is easy to build, and the program is simple and easy to implement; an overlapping area ROI of images can be calculated by the homography matrix between images, and an energy model thereof can be built and solved with a graph cut algorithm; the graph cut algorithm has high time complexity and depends on the number of nodes in a graph; here, images are divided into layers, and solutions are obtained layer by layer and iterated; and finally, a stitched image is further optimized by simple linear fusion of stitching seams and histogram equalization of the stitched image.