Image Stitching with 3D Distortion Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to effectively stitch together images of overlapping scenes with changing camera positions and orientations, particularly in applications like aerial vehicle surveillance, due to the lack of camera position and orientation information.
Innovation Solution
A method that detects feature points and extracts texture descriptors from images, applies multiple image transformations to normalize 3-D distortion, and aligns images using inliers to create a panorama, employing techniques like SIFT and RANSAC for robust matching and transformation calculation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If camera position and orientation change over time, then the ability to capture different views of the scene is improved, but the difficulty of stitching together images to create an overall view increases
Solution Approach 1:
The system performs preliminary actions by detecting feature points and extracting texture descriptors from each image before attempting to stitch them together. This preparatory processing of images, including generating multiple transformed versions with different transformations applied, makes the subsequent alignment and stitching process more effective despite camera position and orientation changes.
2Area of stationary object
If feature points are matched between images with large viewing angle differences, then the coverage of the panorama is improved, but the accuracy of feature point matching deteriorates due to 3-D distortion
Solution Approach 1:
The system applies multiple different transformations to the second image, generating a set of transformed images with varying parameters. By comparing feature points across these transformed versions against the first image, the system can identify matches that are robust to 3-D distortion caused by large viewing angle differences, thereby maintaining matching accuracy while expanding panorama coverage.
Solution Approach 2:
The system uses texture similarity comparison as a feedback mechanism to evaluate and refine feature point matches. By calculating texture similarity between matched feature points and using this information to select inliers that fit the transformation model, the system iteratively improves matching accuracy despite the challenges posed by large viewing angle differences and 3-D distortion.
3Manufacturing precision
If multiple image transformations are applied to compensate for 3-D distortion, then the accuracy of panorama creation is improved, but the computational complexity increases
Solution Approach 1:
The system applies multiple transformations to the second image, which is more than the single transformation that would be minimally required. This excessive action of generating multiple transformed versions provides redundancy that improves panorama creation accuracy by allowing selection of the best-matching transformation, while the computational complexity is managed through efficient processing of the transformed images.
Data Source
AI summary
A system, method and computer program product for relating corresponding points in images with an overlapping scene. An example method includes generating transformed images of a target image using different image transformations for each of transformed images. Texture descriptors are extracted for feature points in the transformed images and a reference image. Matched feature points are identified and inliers from matched feature points are selected. An aligning transformation is generated using the inliers for at least one of the transformed images. A panorama image is created with the target image and reference image after the images are aligned.