Affine Image Alignment for Reliable Biometric Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image merging methods require high accuracy in image alignment, leading to user unfriendliness and inaccurate user identification due to improper alignment of biometric features.
Innovation Solution
A computer-implemented method determines an affine transformation to align features of two images, ensuring accurate alignment by checking for the existence of a valid transformation and using a neural network to calculate parameters, allowing for high-accuracy user identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If image merging methods are applied to combine separate images of a user, then user identification reliability is improved, but alignment accuracy requirements become too strict making the system user-unfriendly
Solution Approach 1:
The patent changes the parameter of transformation model from rigid to affine, allowing for stretching, rotating, and shifting operations. This enables the system to accommodate variations in image quality and alignment while maintaining reliable feature matching, thus improving user-friendliness without sacrificing identification accuracy
Solution Approach 2:
The patent implements a feedback mechanism where the system determines whether an affine transformation exists between images and outputs an indication of validity. This feedback loop allows the system to adapt to different image conditions and provide accurate identification while being forgiving of minor alignment issues, making it user-friendly
2Measurement precision
If high accuracy alignment is required for image merging, then user identification accuracy is improved, but errors increase due to lack of proper alignment
Solution Approach 1:
By changing from rigid transformation to affine transformation with parameters for stretching, rotating, and shifting, the system achieves more robust alignment. This allows accurate feature matching even when images are not perfectly aligned, reducing errors while maintaining high identification accuracy
Solution Approach 2:
The patent replaces traditional rigid alignment mechanisms with a neural network-based affine transformation system. This substitution enables the system to handle misaligned images more gracefully by learning appropriate transformations, thereby reducing alignment-related errors while maintaining precision
3Productivity
If affine transformation parameters are determined without checking existence, then processing speed is improved, but transformation accuracy becomes unreliable
Solution Approach 1:
The patent incorporates a feedback check that determines whether an affine transformation actually exists between the input images. This validation step prevents the system from producing inaccurate transformations, ensuring reliability while maintaining efficient processing by avoiding unnecessary computations on unsuitable image pairs
Data Source
AI summary
A computer-implemented method for determining an affine transformation for transforming a second image so that features of the second image coincide with corresponding features of a first image, the computer-implemented method comprising obtaining the second image comprising the features and obtaining the first image comprising the corresponding features, processing the images by a transformation calculator, thereby determining an indication of whether the affine transformation exists and determining parameters defining the affine transformation, and outputting the affine transformation, comprising outputting the indication of whether the affine transformation exists and outputting a data structure indicative of the parameters defining the affine transformation.