Face and Auxiliary Region Feature Extraction for Non-Frontal Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Face recognition techniques often fail to achieve acceptable recognition accuracy for non-frontal and poorly registered facial images.
Innovation Solution
A method that detects a face region and an auxiliary identification region in an input image, extracts features from both, and generates an indexing datum for storage, enabling improved recognition even with non-frontal and poorly registered images by utilizing facial and auxiliary identification features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional face recognition techniques are used, then recognition accuracy is improved for frontal images, but recognition accuracy deteriorates for non-frontal and poorly registered images
Solution Approach 1:
The patent divides the identification task into two segments: face region analysis and auxiliary identification region analysis. The auxiliary region (separate from the face) is extracted and processed independently to provide additional identification features, allowing the system to maintain accuracy across different facial views and registrations.
Solution Approach 2:
The patent transitions from relying solely on the two-dimensional face region to incorporating an additional dimension of information from the auxiliary identification region. This extra spatial dimension provides complementary features that improve recognition when facial views vary.
2Device complexity
If only face region features are extracted, then processing complexity is reduced, but recognition accuracy deteriorates for non-standard facial views
Solution Approach 1:
The system segments the image processing into distinct regions (face region and auxiliary identification region), allowing each to be processed with appropriate complexity. The auxiliary region provides additional accuracy without requiring complex modifications to the face processing pipeline.
Solution Approach 2:
The auxiliary identification region acts as an intermediary source of identification information. When face region features are insufficient (non-standard views), the auxiliary region provides mediating features that bridge the gap and enable accurate recognition.
3Measurement precision
If auxiliary identification region is incorporated, then recognition accuracy is improved for non-frontal images, but processing time increases
Solution Approach 1:
The auxiliary identification region is identified and extracted in advance as part of the initial image processing pipeline. This preliminary action prepares additional identification features before they are needed, reducing processing delays during recognition operations.
Solution Approach 2:
The system creates a copy of the auxiliary identification region features and stores them separately from the face region processing. This allows parallel processing of face and auxiliary features, minimizing the time penalty for incorporating additional identification information.
Data Source
AI summary
Method, apparatus, and systems for extracting features from face regions and auxiliary identification regions of images are described. In one aspect, a face region containing at least a portion of a person's face is detected in an input image. An auxiliary identification region outside the face region is identified in the input image. Facial features characterizing the face region are determined. Auxiliary identification features characterizing the auxiliary identification region are calculated. An indexing datum is derived from the facial features and the auxiliary identification features. The indexing datum is stored in association with the input image.