Camera Array Image Fusion Using Reference Grid Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for combining images from an array of parallel cameras are inefficient, failing to effectively preserve and enhance details while reducing noise, which is crucial for producing high-resolution images and determining depth information.
Innovation Solution
A method that involves selecting a reference camera, estimating misalignment and disparity parameters, mapping image data onto a reference grid using fractional offsets, and interpolating output pixels to produce a high-resolution image, which can be achieved without perfect camera calibration and in a single interpolation process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If images from an array of cameras are combined using existing methods, then high-resolution images can be produced, but the process is inefficient and fails to effectively preserve and enhance details while reducing noise
Solution Approach 1:
The patent applies preliminary action by performing misalignment estimation and disparity parameter calculation before the actual image fusion process. The method pre-processes the images from multiple cameras to determine their relative positions and orientations, then uses these pre-calculated parameters to guide the fusion process. This preliminary preparation enables more efficient and accurate image combining while preserving details and reducing noise.
Solution Approach 2:
The patent utilizes parameter changes by estimating misalignment parameters (rotation, translation) and disparity values between cameras, then dynamically adjusting the mapping of image data onto the reference grid based on these parameters. The method changes the spatial parameters of each camera's image data to account for their different positions and orientations, enabling accurate fusion despite variations in camera alignment.
2Reliability
If perfect camera calibration is required for efficient image fusion, then image fidelity can be maximized, but the complexity and difficulty of calibration increases significantly
Solution Approach 1:
The patent applies self-service by enabling the system to automatically estimate misalignment and disparity parameters directly from the captured images without requiring external calibration equipment or complex manual calibration procedures. The method uses the image data itself to determine the relative positions and orientations of cameras, allowing the system to self-calibrate and produce high-fidelity fused images without perfect pre-calibration.
Solution Approach 2:
The patent replaces the mechanical calibration system with a computational approach. Instead of using physical calibration objects and manual measurement procedures, the method substitutes these with algorithmic estimation of misalignment parameters and disparity values from the image data itself, significantly reducing the mechanical complexity while maintaining or improving image fidelity.
3Manufacturing precision
If multiple processing steps are used to combine images from camera arrays, then detail preservation may improve, but the processing time and computational complexity increases
Solution Approach 1:
The patent merges multiple processing steps into a unified image fusion framework. By combining misalignment correction, disparity estimation, and image fusion into a single integrated process that maps all camera images onto a common reference grid simultaneously, the method achieves detailed image fusion without the sequential processing delays of separate steps, reducing overall processing time while maintaining detail enhancement.
Data Source
AI summary
An image fusing method, apparatus and system for fusing images from an array of cameras, method includes selecting a camera from the array of cameras as a reference camera, estimating misalignment between the retrieved input images from the reference camera and the retrieved input images from the other cameras, estimating misalignment parameters between the reference camera and the other cameras, estimating local disparity between the reference camera image data and the other cameras based on the estimated misalignment parameters, using the estimated misalignment parameters and the estimated disparity values, mapping the image data into a reference camera grid, the retrieved input image data from the other cameras in the array of cameras is fused in the reference camera grid utilizing fractional offsets from integer coordinates, and producing an output image grid on the reference camera grid and interpolate output pixels using processed data for producing a high resolution image.