A robust face super-resolution algorithm and its application in low-resolution face recognition system
- PDF / 3,342,088 Bytes
- 26 Pages / 439.642 x 666.49 pts Page_size
- 49 Downloads / 207 Views
A robust face super-resolution algorithm and its application in low-resolution face recognition system Shyam Singh Rajput1
· K. V. Arya2
Received: 26 May 2019 / Revised: 11 March 2020 / Accepted: 15 May 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract In real-world surveillance scenario, the face recognition (FR) systems pose a lot of challenges due to the captured low-resolution (LR) and noisy probe images. A new face super-resolution (SR) algorithm is proposed to design a recognition model overcoming the challenges of existing FR systems. The proposed SR algorithm inherits the merits of functional-interpolation and dictionary-based SR techniques. The functional interpolation assists in generating more discriminable output, whereas the dictionary-based approach assists in eliminating noise effects from the reconstruction process. Consequently, it produces more discriminable and noise-free high-resolution (HR) images from captured noisy LR probe images, suitable for real-world problems like low-resolution face recognition. The results obtained from the experiments performed on several popular face image datasets including FEI, FERET, and CAS-PEAL-R1 show that the proposed algorithm performs better than all the comparative SR methods. Keywords Low-resolution face recognition · Super-resolution · Noisy face images · Functional-interpolation · Dictionary or Training based models
1 Introduction In last three-four decades, numerous systems have been invented for recognition of face images. These systems perform admirably under several assumptions like the size of the detected face is large enough, and carries adequate information for classification. However, in real-time scenario (i.e., uncontrolled environment), these assumptions may not be Shyam Singh Rajput
[email protected] K. V. Arya [email protected] 1
Department of Computer Science & Engineering, National Institute of Technology, Patna-800005, India
2
ABV-Indian Institute of Information Technology and Management, Gwalior- 474015, India
Multimedia Tools and Applications
valid due to various reasons such as cheap and low-resolution imaging devices [34], processing and restoration error [13, 27], and the higher distance of an object from the device [33]. Therefore, existing face recognition (FR) systems perform poorly in uncontrolled environment. This is known as low-resolution face recognition (LR-FR) problem [52]. Several methods proposed in recent years to solve the LR-FR problem which can broadly be categorized into two parts [17] (i) unified feature space based [2, 17, 24, 48], and (ii) superresolution based methods [9, 52]. FR process using these two techniques is illustrated in Fig. 1. With the aim of resolving dimensionality mismatch problem between low-resolution (LR) probe images and high-resolution (HR) gallery images, the idea of matching in unified feature space was originally introduced by Li et al. [24]. In this, they project the features of LR probe and HR gallery images in unified feature space
Data Loading...