Sparse/low-rank analytics and representation have been emerging as important topics for robust image processing and visual representation due to their great success to image restoration, de-noising and classification, etc. Although recent decade has witnessed lots of efforts on the study of sparse/low-rank analytics, and significant progress were made to improve the representation ability, some issues still remain unsolved.
For example, sparse/low-rank representation algorithms usually utilize the single-layer structures, so they fail to obtain the deep representations with more useful and valuable hidden hierarchical information discovered. With the fast development of deep learning and deep neural networks, it will be helpful to propose the deep/multi-layer sparse and low-rank representation frameworks for robust visual representation. It is known that both deep learning and sparse/low-rank coding are powerful representation learning systems based on different mechanisms and principles, but how to integrate them to improve the performance is still unclear and noteworthy exploring, which is the main goal of this special issue.
Although certain efforts have been made to incorporate the deep neural networks into sparse/low-rank analytics, most designs of so-called deep frameworks are still less straightforward. For example, some algorithms use deep features of deep models for sparse/low-rank analytics, or perform the sparse/low-rank analytics firstly and use the recovered data for deep learning. Although certain deep features or representations can be obtained by this kind of deep sparse/low-rank analytics, they still suffer from some drawbacks.
For example, they only simply add together multiple shallow sparse/low-rank coding layers, so current models still cannot produce accurate representations of visual data. Thus, it is now necessary to explore advanced integrated deep sparse/low-rank coding algorithms and theories for robust visual representation.
In this special issue, we solicit original research papers from diverse research fields, developing new deep sparse/low-rank analytics model for representing and understanding visual data, which aims to reduce the gap between sparse/low-rank coding and deep learning.
The topics of interest include, but are not limited to:
Survey/vision/review of sparse/low-rank visual analytics
Deep/multi-layer sparse coding or low-rank coding
Relations between sparse/low-rank coding and deep learning
Deep representation learning
Deep sparse or low-rank coding neural network
Convolutional sparse/low-rank coding
Robust sparse/low-rank subspace discovery
Theory and optimization for deep representation learning
Applications to robust image processing (e.g., restoration and de-noising) and recognition
Paper submission due: October 20, 2020
First notification: November 20, 2020
Final decision made on all manuscripts: March 20, 2021
Managing Guest Editor:
Prof. Zhao Zhang, School of Computer and Information (School of Artificial Intelligence), Hefei University of Technology, China
Other Guest Editors:
Dr. Paris Giampouras, Mathematical Institute of Data Science, Johns Hopkins University, USA
Dr. Sheng Li, Department of Computer Science, University of Georgia, USA
Prof. Shuicheng Yan, Department of Electrical and Computer Engineering, National University of Singapore, Singapore