Abstract
Image set matching (ISM) has attracted increasing attention in the field of computer vision and pattern recognition. Some studies attempt to model query and gallery sets under a joint or collaborative representation framework, achieving impressive performance. However, existing models consider only the competition and collaboration among gallery sets, neglecting the inter-instance relationships within the query set which are also regarded as one important clue for ISM. In this paper, inter-instance relationships within the query set are explored for robust image set matching. Specifically, we propose to represent the query set instances jointly via a combined dictionary learned from the gallery sets. To explore the commonality and variations within the query set simultaneously to benefit the matching, both low rank and class-level sparsity constraints are imposed on the representation coefficients. Then, to deal with nonlinear data in real scenarios, the‘kernelized version is also proposed. Moreover, to tackle the gross corruptions mixed in the query set, the proposed model is extended for robust ISM. The optimization problems are solved efficiently by employing singular value thresholding and block soft thresholding operators in an alternating direction manner. Experiments on five public datasets demonstrate the effectiveness of the proposed method, comparing favorably with state-of-the-art methods.
Highlights
Image set matching (ISM) or image set classification, which regards one set of images as a sample, has recently attracted considerable attention due to its widespread applications such as video-based face recognition, multi-view object recognition, dynamic scene classification [1,2,3,4,5,6,7,8,9,10]
We proposed a joint representation model with both low rank and class-level constraints imposed on the representation coefficients to explore fully the inter-instance relationships within the query set for improving image set matching
Compared with joint representation models with other sparsity constraints, the model with class-level sparsity constraint is more appropriate for image set matching, due to its enhanced inter-class discrimination by class label prior
Summary
Image set matching (ISM) or image set classification, which regards one set of images as a sample, has recently attracted considerable attention due to its widespread applications such as video-based face recognition, multi-view object recognition, dynamic scene classification [1,2,3,4,5,6,7,8,9,10]. SRC or CRC implicitly describes the relationships (competition and collaboration) among the training samples while measuring directly the distances/similarities between testing and training samples Such methods are not sensitive to the feature extraction techniques, i.e., whatever kind of extracted feature is fed to the models based on SRC or CRC leads to good performances. Motivated by the above insights, in this paper, we propose a new joint representation model to highlight the inter-instance relationships within the query set for improving ISM. A joint sparse representation model with class-level sparsity constraint is chosen for ISM problem and a low rank regularization is added to reveal thoroughly the intra-set and inter-set relationships to improve the ISM performance.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have