Image Stitching Quality Assessment Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image stitching processes in multi-camera systems often introduce distortions due to misalignment of pixels, leading to noticeable discontinuities and degraded image quality, necessitating a consistent assessment of image quality to adjust stitching parameters effectively.
Innovation Solution
A system that uses a machine learning module trained on image portions from single and stitched images to assess the quality of stitching by inputting image portions from stitching boundaries, allowing for the selection of optimal stitching parameters to improve image quality, and correlates with subjective human assessments for accurate feedback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If image stitching is performed to create larger field of view images, then the field of view is expanded, but stitching distortions and discontinuities are introduced that degrade image quality
Solution Approach 1:
The system uses a machine learning module to assess stitching quality by analyzing image portions at stitching boundaries, providing feedback that enables selection of optimal stitching parameters to minimize distortions while maintaining expanded field of view
Solution Approach 2:
The system selects parameters for the stitching algorithm based on scores from the machine learning module, dynamically adjusting stitching parameters to optimize the balance between field of view expansion and image quality preservation
2Productivity
If traditional image stitching algorithms are used, then processing speed is maintained, but accurate assessment of stitching quality is difficult to achieve
Solution Approach 1:
The system replaces traditional manual or rule-based quality assessment methods with a machine learning module that automatically evaluates stitching quality, achieving both accuracy and processing efficiency through automated intelligent assessment
3Manufacturing precision
If stitching parameters are adjusted to reduce distortions, then image quality is improved, but the complexity of parameter selection and optimization increases
Solution Approach 1:
The machine learning module autonomously assesses stitching quality and the system automatically selects optimal parameters based on assessment scores, eliminating the need for manual parameter tuning and reducing operational complexity while maintaining high image quality
Data Source
AI summary
Systems and methods are disclosed for image signal processing. For example, methods may include receiving a first image from a first image sensor; receiving a second image from a second image sensor; stitching the first image and the second image to obtain a stitched image; identifying an image portion of the stitched image that is positioned on a stitching boundary of the stitched image; and inputting the image portion to a machine learning module to obtain a score, wherein the machine learning module has been trained using training data that included image portions labeled to reflect an absence of stitching and image portions labeled to reflect a presence of stitching, wherein the image portions labeled to reflect a presence of stitching included stitching.