Face Orientation Detection Using Regional Characteristic Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face orientation detection methods from captured images require increased calculation and duplication of processes when creating multiple inclined face images and calculating similarity, making them inefficient.
Innovation Solution
A method that determines a face region within an image, sets adjacent detecting regions, acquires characteristic amounts from these regions, and compares them to judge the orientation, reducing the complexity of the process and enabling accurate detection regardless of image differences or background similarities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple inclined face images are created and similarity calculation is performed, then face orientation detection accuracy is improved, but calculation amount increases and process complexity increases
Solution Approach 1:
The patent extracts only the necessary characteristic information (characteristic amounts) from the face image that is sufficient for orientation detection, rather than performing complete similarity calculations on multiple inclined face images. This extraction approach maintains detection accuracy while significantly reducing computational complexity.
Solution Approach 2:
The patent segments the face detection process into distinct regions (first detecting region, second detecting region, third detecting region) and analyzes characteristic amounts from each region separately. This segmentation allows for simplified comparison operations while maintaining accurate orientation detection through regional analysis.
2Measurement precision
If multiple inclined face images are created and similarity calculation is performed, then face orientation detection accuracy is improved, but calculation amount increases
Solution Approach 1:
The patent extracts only the essential characteristic amounts from segmented face regions that are necessary for orientation determination, avoiding the computationally intensive task of calculating complete similarity metrics for multiple inclined face images. This extraction strategy maintains detection precision while reducing calculation volume.
Solution Approach 2:
The patent performs partial analysis by examining characteristic amounts from specific detecting regions rather than analyzing the entire face image comprehensively. This partial action approach provides sufficient information for orientation detection without the excessive calculation required by complete similarity analysis.
3Measurement precision
If characteristic amounts are acquired with reference to color component, then detection accuracy for different skin tones is improved, but processing complexity increases
Solution Approach 1:
The patent applies local quality by using color component reference values that are specific to each detecting region. Each region has its own characteristic color reference that accounts for variations in skin tone and lighting conditions, enabling accurate detection across different skin tones while maintaining relatively simple processing through localized rather than global analysis.
Data Source
AI summary
In order to detect an orientation of a human face included in an image consisted of a plurality of pixels, a face region is determined within the image so as to include the human face. A plurality of detecting regions are set so as to be adjacent to the face region. A plurality of first amounts are acquired. Each of the first amount is characteristic to one of the detecting regions. The first amounts are compared to each other to judge the orientation.