Facial Feature Point Position Correction via Reliability Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting facial feature points in images face challenges when erroneous reliability maps are generated due to factors like changes in illumination or obstruction, leading to incorrect positioning of facial feature points, which can affect the accuracy of face direction estimation and authentication systems.
Innovation Solution
A facial feature point position correcting device and method that calculates initial positions based on reliability maps, judges off-position points using a statistical face shape model, and corrects positions by calculating differences according to a prescribed evaluation function, ensuring high-accuracy output even with low-reliability input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If facial feature point positions are detected using reliability maps generated from images with illumination changes or obstructions, then initial detection can be performed, but erroneous reliability maps cause incorrect positioning and shift other feature point positions
Solution Approach 1:
The patent applies feedback by using the detected facial feature point positions to regenerate reliability maps and iteratively refine the detection results. The system feeds back the initial detection results to correct erroneous reliability maps, which then improve subsequent detection accuracy, creating a closed-loop system that resolves the contradiction between initial detection capability and reliability map accuracy.
Solution Approach 2:
The patent performs preliminary action by generating initial facial feature point positions and reliability maps before final detection. This preliminary detection phase allows the system to identify and correct erroneous reliability maps in advance, preventing them from affecting final position accuracy and enabling more reliable subsequent detection.
2Reliability
If penalty is assigned to facial feature point positions far from statistical face shape model positions, then plausible positions can be acquired even with hidden features, but erroneous reliability maps cause positions to be shifted to average positions
Solution Approach 1:
The system uses feedback to detect when penalty-based correction shifts positions due to erroneous reliability maps. By feeding back the corrected positions to regenerate reliability maps, the system identifies and corrects the erroneous maps, ensuring that penalty assignment restores positions to accurate locations rather than shifted average positions.
Solution Approach 2:
The patent performs preliminary detection to establish accurate baseline positions before applying penalty-based correction. This preliminary action ensures that the statistical face shape model comparison is based on accurate initial positions, preventing erroneous reliability maps from causing shifts to incorrect average positions during penalty assignment.
3Measurement precision
If geometric constraint process is executed to correct positions based on positional relationships, then positions can be corrected when normalization errors occur, but it is impossible to judge which positions are deviated from standard face shape
Solution Approach 1:
The patent uses feedback to automatically identify which facial feature point positions are deviated from standard face shape. By feeding back the detected positions to compare against the statistical face shape model and reliability maps, the system automatically judges which positions require correction, eliminating the need for complex manual judgment while maintaining high precision.
Solution Approach 2:
The system applies self-service by automatically identifying and correcting its own position deviations using the statistical face shape model and reliability maps. The facial feature point position correcting device performs self-diagnosis and self-correction, determining which positions are deviated and correcting them without external intervention, thereby reducing process complexity while maintaining accuracy.
Data Source
AI summary
Facial feature point reliability generating means generates a reliability map of each facial feature point from a facial image. Initial facial feature point position calculating means calculates the position of each facial feature point in the facial image based on the reliability map. Off-position facial feature point judgment means judges whether or not each facial feature point is an off-position facial feature point not satisfying a prescribed condition. Facial feature point difference calculating means calculates the difference between the position of each facial feature point, excluding those judged as the off-position facial feature points, and the position of a corresponding point of the facial feature point. Facial feature point position correcting means corrects the determined positions of the facial feature points based on the results of the judgment by the off-position facial feature point judgment means and the calculation by the facial feature point difference calculating means.