Facial Recognition via Two-Stage Landmark Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Facial recognition systems face decreased recognition rates when there are significant changes in pose or size of facial landmarks, leading to inaccuracies in identifying individuals.
Innovation Solution
A method and apparatus for facial recognition that involves detecting initial landmarks, normalizing the facial image, updating the position of intermediate landmarks based on confidence calculations using probabilistic models, and subsequent renormalization to improve feature extraction and recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If facial recognition is performed based on front face only, then recognition accuracy is maintained for standard poses, but recognition rate decreases when user pose changes substantially
Solution Approach 1:
The patent transforms the facial image through geometric transformation parameters (rotation, scaling, translation) to align the detected pose with the front face orientation. By dynamically adjusting transformation parameters based on detected landmark positions, the system maintains recognition accuracy across various poses while adapting to pose changes.
Solution Approach 2:
The patent performs preliminary pose estimation and normalization before the actual recognition process. By detecting initial landmarks and estimating the transformation required to convert the current pose to front face orientation in advance, the system prepares the image data to ensure accurate recognition regardless of the original pose.
2Productivity
If single normalization is performed using initial landmarks, then processing speed is maintained, but recognition accuracy decreases when landmarks have large pose or size changes
Solution Approach 1:
The patent divides the normalization process into multiple stages: first normalization using initial landmarks, then intermediate landmark detection and updating, followed by second normalization. This segmented approach improves accuracy for challenging cases while maintaining efficiency through selective processing - the second normalization is applied based on whether pose/size changes exceed thresholds.
Solution Approach 2:
The patent applies the second normalization selectively rather than to all images. When intermediate landmarks show significant position changes (exceeding thresholds), the system performs the additional normalization step. This partial application of excessive action ensures high accuracy for difficult cases while avoiding unnecessary processing for already well-aligned images.
3Reliability
If intermediate landmark positions are updated using confidence-based methods, then feature extraction robustness is improved, but computational complexity increases
Solution Approach 1:
The patent replaces complex iterative optimization methods with a confidence-based update mechanism. Instead of using computationally intensive mechanical optimization algorithms to refine landmark positions, the system uses probabilistic confidence values derived from image data to directly update positions, achieving robust feature extraction with reduced computational complexity.
Solution Approach 2:
The patent introduces confidence values as an intermediary between raw landmark detection and final position determination. These confidence values, calculated from image data and detection quality metrics, serve as a mediator to guide landmark position updates, improving reliability without requiring complex optimization systems.
Data Source
AI summary
At least some example embodiments disclose a method and apparatus for facial recognition. The facial recognition method includes detecting initial landmarks from a facial image, first normalizing the facial image using the initial landmarks, updating a position of at least one of intermediate landmarks based on the first normalizing, the intermediate landmarks being landmarks transformed from the initial landmarks through the first normalizing, second normalizing the facial image after the updating and recognizing a face using a feature of the second normalized facial image.