This code includes the detailed implementation of the paper:
Reference:
Liu, Q., Lai, Z., Zhou, Z., Kuang, F., & Jin, Z. (2016). A truncated
nuclear norm regularization method based on weighted residual error for
matrix completion. IEEE Transactions on Image Processing, 25(1), 316-330.
It is partially composed of TNNR code implementation.
Reference:
Hu, Y., Zhang, D., Ye, J., Li, X., & He, X. (2013). Fast and accurate matrix
completion via truncated nuclear norm regularization. IEEE Transactions on
Pattern Analysis and Machine Intelligence, 35(9), 2117-2130.
The code contains:
|--------------
|-- TNNR_WRE_main.m entrance to start real image experiment
|-- TNNR_WRE_synthetic.m entrance to start synthetic experiment
|-- function/ functions of TNNR-WRE algorithm
|-- PSNR.m compute the PSNR and Erec for recovered image
|-- TNNR_WRE_algorithm.m main part of TNNR-WRE implementation
|-- weight_matrix.m compute weight matrix in an increasing order
|-- weight_sort.m sort the sequence of weight value according to
observed elements; rows with more observed
elements are given smaller weights
|-- image/ directory for original images
|-- mask/ directory for various mask types, 300x300
|-- result/ directory for saving experimental results
|-------------
For algorithm interpretation, please read Liu et al. (2016) paper and Hu et
al. (2013) paper, in which more details are demonstrated.
If you have any questions about this implementation, please do not hesitate
to contact me.
Ph.D. Candidate, Shengke Xue,
College of Information Science and Electronic Engineering,
Zhejiang University, Hangzhou, P. R. China,
e-mail: (either one is o.k.)
xueshengke@zju.edu.cn, xueshengke1993@gmail.com.