Blind Video Stream Registration Using Calibration LUTs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Registering images captured by spectrally and spatially disparate cameras is computationally complex and time-consuming, particularly for video images, and existing techniques like feature-based and intensity-based registration are ineffective, complicating the process.
Innovation Solution
A processor-based apparatus and method for real-time, blind registration of spatially and spectrally disparate video images, using look-up tables (LUTs) configured during offline calibration to perform resizing, distortion correction, and resampling without requiring feature or content knowledge of the images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature-based or intensity-based registration techniques are used to register spectrally and spatially disparate images, then registration accuracy may be improved, but computational complexity and processing time increase significantly
Solution Approach 1:
The patent applies preliminary action by performing offline calibration to generate lookup tables (LUTs) that contain pre-computed registration parameters. During real-time operation, the system simply applies these pre-computed transformations without performing complex feature matching or intensity optimization, thus resolving the contradiction between registration accuracy and computational complexity.
2Productivity
If real-time registration is implemented for video images from spectrally disparate cameras, then processing speed is improved, but computational resources required increase
Solution Approach 1:
The system performs computationally intensive calibration operations offline to generate LUTs, then applies these pre-computed transformations in real-time with minimal computational resources. This separates the heavy computation from the real-time processing requirement, allowing fast video registration without excessive computational resource consumption during operation.
3Loss of time
If images are registered without using feature or content knowledge (blind registration), then processing time is reduced for real-time application, but registration precision may be compromised
Solution Approach 1:
The patent performs preliminary calibration using known geometric relationships and captured images to compute transformation parameters that are stored in LUTs. During real-time blind registration, these pre-computed parameters enable accurate spatial alignment without needing to analyze image features or content, thus achieving both speed and precision.
Data Source
AI summary
A first image of a scene captured by a first camera is received. The first image is resized to have an instantaneous field of view (IFoV) corresponding to the IFoV of a second image of the scene captured by a second camera. A distortion-corrected resized first image is constructed. A resampled distortion-corrected, resized first image is constructed to provide a resampled, distortion-corrected, resized first image, thereby registering first image to the second image.