Face recognition using part-based dense sampling local features
Introduction
Over the last few decades, face recognition has been an active research topic in computer vision and pattern recognition, and a lot of such systems emerged across various applications, such as video surveillance, access control, image retrieval and automatic log-on for personal computers or mobile devices. Face recognition is an easy task for human visual system, but it is quite challenging for automatic face recognition system due to the dramatic variations among the appearance of the same subject, which is caused by a lot of factors such as image resolution, illumination, expression, pose, occlusion, etc. These various visual complications deteriorate the performance of a face recognition system dramatically.
Automatic face recognition mainly involves three stages: detection, representation and classification. In the detection stage, the face is localized in an image. The representation stage refers to extracting specific features from a given face image. The classification stage deals with the final decisive process whether an unknown face image can be categorized into a target subject. Among the three main parts, face representation occupies a crucial position of a face recognition system. A critical point to the performance of a face recognition system is the face representation׳s discriminative power for various individuals and invariance to external environment changes, such as resolution and poses variation. A robust and effective face representation is indispensable to performance improvement.
Traditional face representations such as Local Binary Pattern (LBP) [1,2] and Gabor [3], [4] achieved good performance on face images with good quality. LBP is good at coding fine details of facial appearance and texture, and Gabor features encode facial shape and appearance over a range of coarser scales. A fusion strategy of LBP and Gabor can obtain promising performance for face recognition [20], [22], [23], [24]. But this strategy has the problem of complex computations. Meanwhile, the fusion of LBP and Gabor is sensitive to resolution variation.
In addition to efficient face representation, the performance of face recognition system relies on the metric techniques used to measure the similarity of two face representations. The recent frameworks mostly learn a dataset specific metric during the training stage, which can result in the improvement of the recognition performance. It is mainly because it learns the bias of the dataset and thus adapts to the specific distribution. In this paper, we show that our face representation can be used with a very simple face matching method that does not require complex metric learning algorithm. Once features are extracted from face images, we first use whitened Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) to project them into a low dimensional space, and then apply cosine similarity on the more compact features to measure the similarity between image pairs.
We experimentally evaluated our proposed face representation on two databases. The first one, which is composed of a subject’s high resolution identification (ID) photo and relative low resolution driver׳s license photo, is collected by us. The second one is the public FERET database [5]. Experimental results show that our method has very good generalization capability and outperforms many traditional methods both in performance and efficiency.
The main contributions of this paper are summarized as follows:
- 1.
We propose a novel face representation strategy for face recognition, which is effective and robust to resolution variation.
- 2.
Experimental results show that the proposed face representation can significantly improve face recognition performance without time consuming Gabor convolution.
The rest of the paper is organized as follows: Section 2 reviews the related work of face recognition. Section 3 describes details of the whole framework and focuses on the proposed new face representation strategy. Section 4 shows experimental results and Section 5 gives conclusion of this paper.
Section snippets
Related work
Recent years, advances in the face representation have been a major source of progress in the field of face recognition. Many feature extraction methods have been introduced to characterize face images for various specific tasks, including holistic and local features.
The common goal of these holistic feature extraction methods is to learn a compact and low-dimensional feature subspace for face representation, so that the intrinsic characteristics of face images could be preserved. Among these
Proposed algorithm for face verification
In this section, we describe the algorithm proposed for face representation. We start from detecting facial feature points with Intraface system [25] on face images, extracting patches with fixed size centered on the detected feature points from face images over different scales, and applying dense sampling strategy to get a fixed number of blocks on each patch. Then on each block we extract histograms of a new proposed local feature, after that PCA+LDA are applied to generate compact
Experimental results and analysis
To evaluate performance of our proposed algorithm on face recognition problem, we applied it on a database consisting of face image pairs on identification (ID) card and driver׳s license. The results were compared with some traditional methods, such as LBP and LBP+Gabor, to show the effectiveness of our algorithm. We also applied our proposed algorithm on FERET [5], and compared it with some state-of-the-art methods.
Conclusion
The study has shown that using local feature descriptor in a dense sampling way yields good results for face recognition with resolution variations. Experimental results showed two main merits of our method: high recognition rate and fast feature extraction speed. The high recognition rate is mainly due to the LDF feature extraction method and the dense sampling strategy. The quick feature extraction speed springs from the omitting Gabor filter which is widely used for face recognition. As
Acknowledgment
The work is partially supported Shenzhen Oversea High Talent Innovative fund (Grant no. KQCX20140521161756231), the Natural Science Foundation of China (NSFC) (Grant No. 61527808), Guangdong Natural Science Foundation (Grant No. S2013010016601), Shenzhen Municipal Science and Technology Plan Projects (Grant No. JCYJ20130401095947234) and the CICAEET fund and the PAPD fund.
Jiaqi Zhang is studying for his master degree from Tsinghua University since 2013. His general interests lie in face recognition, machine learning and deep learning.
References (30)
- et al.
A new LDA-based face recognition system which can solve the small sample size problem
Pattern Recognit.
(2000) - et al.
A direct LDA algorithm for high-dimensional data—with application to face recognition
Pattern Recognit.
(2001) - et al.
Multiresolution gray-scale and rotation invariant texture classification with local binary patterns
Pattern Anal. Mach. Intell. IEEE Trans.
(2002) - et al.
Face recognition with local binary patterns. Computer vision-eccv 2004
(2004) - et al.
Gabor feature based classification using the enhanced fisher linear discriminant model for face recognition
Image Process. IEEE Trans.
(2002) Gabor-based kernel PCA with fractional power polynomial models for face recognition
Pattern Anal. Mach. Intell. IEEE Trans.
(2004)- et al.
The FERET evaluation methodology for face-recognition algorithms
Pattern Anal. Mach. Intell. IEEE Trans.
(2000) - et al.
Eigenfaces for recognition
J. Cognit. Neurosci.
(1991) - et al.
Eigenfaces vs. fisherfaces: recognition using class specific linear projection
Pattern Anal. Mach. Intell. IEEE Trans.
(1997) - et al.
Face recognition using Laplacianfaces
Pattern Anal. Mach. Intell. IEEE Trans.
(2005)
Graph embedding and extensions: a general framework for dimensionality reduction
Pattern Anal. Mach. Intell. IEEE Trans.
MPCA: multilinear principal component analysis of tensor objects
Neural Netw. IEEE Trans.
A doubly weighted approach for appearance-based subspace learning methods
Inf. Forensics Secur. IEEE Trans.
KPCA plus LDA: a complete kernel Fisher discriminant framework for feature extraction and recognition
Pattern Anal. Mach. Intell. IEEE Trans.
Robust face recognition via sparse representation
Pattern Anal. Mach. Intell. IEEE Trans.
Cited by (28)
A new synthetic feature generation scheme based on artificial immune systems for robust offline signature verification
2023, Expert Systems with ApplicationsA transformer-based low-resolution face recognition method via on-and-offline knowledge distillation
2022, NeurocomputingCitation Excerpt :Face recognition has been a hot topic for researchers in the field of computer vision [1–6].
Multiple feature descriptors based model for individual identification in group photos
2019, Journal of King Saud University - Computer and Information SciencesCitation Excerpt :The fusion process can be applied at filtration stage, feature generation stage or at recognition stage to improve the accuracy and efficiency of facial recognition. Zhang et al. (2016) used same strategy at feature generation stage using LBP (Local Binary Pattern) and Gabor filters. The combined method provided the robustness against textural variation, structural variation and geometric variation in facial images.
Pose-robust face recognition with Huffman-LBP enhanced by Divide-and-Rule strategy
2018, Pattern RecognitionCitation Excerpt :Therefore, it is quite obvious that the face images of different poses should be treated specifically. As we know, image patches have become a popular way to represent images recently [10,42]. In some sense, image patches are optimal when the purpose is to find the correspondences between two samples with appearance variations.
Patch-based face recognition using a hierarchical multi-label matcher
2018, Image and Vision ComputingCitation Excerpt :Lu et al. [33] developed the Compact Binary Face Descriptor (CBFD) which learns binary codes by removing the redundancy information with unsupervised learning. Zhang et al. [34] proposed a resolution-variance robust representation strategy based on LBP and Gabor features. PCA and LDA are used to reduce feature dimension.
SVM-based writer retrieval system in handwritten document images
2022, Multimedia Tools and Applications
Jiaqi Zhang is studying for his master degree from Tsinghua University since 2013. His general interests lie in face recognition, machine learning and deep learning.
Yao Deng received his master degree from Tsinghua University in 2014. He joined the M-Team in Alibaba Group as an algorithm engineer in 2014. He has worked on face recognition, image retrieval, object and pedestrian detection. His general interests lie in machine learning and pattern recognition, and their applications to Internet.
Zhenhua Guo received the M.S. and Ph.D degree in computer science from Harbin Institute of Technology and the Hong Kong Polytechnic University in 2004 and 2010, respectively. Since April 2010, he has been worked in Graduate School at Shenzhen, Tsinghua University. His research interests include pattern recognition, texture classification, biometrics, video surveillance, etc.
Youbin Chen received the M.S. and Ph.D. degree in Electronic Engineering from University of Science and Technology of China and the Tsinghua University in 1993 and 1997 respectively. Since 2011, he has been worked in Huazhong University of Science and Technology. His research interests include Pattern Recognition, Document Image Analysis, Biometrics, Video Analysis, Machine Learning, Data Mining, Web Mining, etc.