Multi-View Face Descriptor Synthesis via 3D Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional face recognition systems face challenges in dealing with pose changes and illumination variations, leading to decreased accuracy when using single-view query images and requiring multiple query images, and existing 3-D face descriptors are expensive and ineffective for multi-view face images.
Innovation Solution
A method for constructing a multi-view face database and generating a multi-view face descriptor that captures face images from various viewpoints, using an image capturer, eye position determiner, face localization unit, and mosaic view generator to create a mosaic view, and a basis matrix generator, feature extractor, and face descriptor generator to synthesize feature vectors across different viewpoints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a single viewpoint query image is used for face recognition, then the system is simple to operate, but recognition accuracy drops when face images in the database have large pose variations
Solution Approach 1:
The patent transforms the face recognition problem from 2-D image space to 3-D shape space by constructing a three-dimensional face model from multiple 2-D images. This dimensional transformation allows the system to represent faces in a pose-invariant manner, enabling accurate recognition even when query images have different poses from database images.
Solution Approach 2:
The system performs preliminary construction of a 3-D face model and multi-view database before actual recognition. By pre-processing and storing face information in a pose-invariant 3-D representation, the system eliminates the need for multiple query images during operation while maintaining high recognition accuracy across various poses.
2Measurement precision
If multiple query images are used to improve recognition accuracy under pose variations, then recognition performance improves, but the complexity of the system increases
Solution Approach 1:
The patent creates a 3-D digital copy of the face from multiple 2-D images. This 3-D model serves as a pose-invariant representation that can be matched against query images regardless of pose. The copying process transforms complex multi-view matching into simple 3-D model comparison, reducing system complexity while improving accuracy.
3Measurement precision
If conventional 3-D face descriptors are generated using specialized equipment or multiple cameras, then pose changes can be handled, but the cost and complexity of the system increases
Solution Approach 1:
The patent replaces complex mechanical multi-camera acquisition systems with a simpler alternative: capturing multiple 2-D images using standard imaging equipment and computationally reconstructing the 3-D face model. This substitution of mechanical complexity with computational processing achieves the same pose-invariant recognition capability at lower hardware cost and system complexity.
Data Source
AI summary
An apparatus for generating a multi-view face descriptor includes a multi-view face database storing sets of training face images and sets of test face images, each set including a mosaic view of a single identity; a basis matrix generator generating a basis matrix of the training face images shown from respective viewpoints searched from the multi-view face database; a feature extractor extracting feature vectors of a test face image shown from each viewpoint using the basis matrix of the training face images searched from the multi-view face database; a to-be-registered view selector searching the test face images using the feature vectors extracted by the feature extractor and selecting a plurality of viewpoints as registered viewpoints according to the search results; and a face descriptor generator generating a multi-view face descriptor by synthesizing face descriptors having the plurality of selected viewpoints.