Image set face recognition based on extended low rank recovery and collaborative representation
- PDF / 1,717,011 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 27 Downloads / 232 Views
ORIGINAL ARTICLE
Image set face recognition based on extended low rank recovery and collaborative representation Zhanjie Song1,2 · Kaiyan Cui1 · Guangtao Cheng2 Received: 10 March 2018 / Accepted: 18 February 2019 © Springer-Verlag GmbH Germany, part of Springer Nature 2019
Abstract In the real-world face recognition problems, the collected query set images often suffer serious disturbances. To address the problem, we propose an image set face recognition method based on extended low rank recovery and collaborative representation. By exploiting a Frobenius norm term, an extended low rank representation model is firstly developed to remove all possible disturbances from the query set and reconstruct the rank-one query set. To improve the computational efficiency, a compact and discriminative dictionary is learned from the large gallery set, and the closed form solutions for both the dictionary atom and the coding coefficient are straightway derived. The final classification is performed by using any frame in the reconstructed query set instead of using the whole set, which can further improve the running efficiency. Extensive experiments are conducted on the benchmark Honda/USCD and Youtube Celebrities database to verify that the proposed method outperforms significantly the state-of-the-art methods in terms of robustness and efficiency. Keywords Image set · Low rank representation · Sparse representation · Face recognition · Image denoising
1 Introduction With the rapid development of digital imaging and communication techniques, image sets can be easily collected from multi-view images using multiple cameras [1], long term observations [2], personal albums and news pictures [3] etc.. Therefore, image set based face recognition (ISFR) has attracted great research interest during the past two decades [4–30]. ISFR focuses on researching how to model a set and consequently how to compute the distance/similarity between each of the gallery sets (image sets that are used to learn a face recognition algorithm) and the query set (an unknown set of images that are to be recognized by the learned algorithm). Relevant previous approaches to set classification can be broadly partitioned into parametric and nonparametric approaches. Parametric methods model each image set as a parametric distribution, and use KullbackLeibler divergence to measure the similarity between the * Kaiyan Cui [email protected] 1
School of Mathematics, Tianjin University, Tianjin 300354, China
Visual Pattern Analysis Research Lab, Tianjin University, Tianjin 300072, China
2
distributions [2, 31]. The disadvantage of parametric set modeling lies in the difficulty of parameter estimation under limited training data, and it may fail when the real gallery and query sets are not fitted by the estimated parametric model [2, 4, 6]. Therefore, many nonparametric methods have also been applied to measure the similarity between two image sets [4–11, 32]. This is certainly useful in many computer vision applications where the data acquisition co
Data Loading...
