This code is the detailed implementation of A Novel Tensor Factorization-Based Method with Robustness to Inaccurate Rank Estimation.
TCDLR_toy_demo.m is the implement of TCDLR algorithm without rank estimation for toy data with different size;
TCDLR_RE_img_demo.m is the implement of TCDLR-RE algorithm for Berkeley Segmentation Datase(BSD)[1]. data/small_img.mat contains 50 random images extracted for the BSD.
matrix-and-tensor-completion-algorithm/TCDLR_RE_background_modeling.m is the implement of TCDLR-RE algorithm for background modeling task. We use the BMnet-2016 dataset [2] to evaluate our algorithm.
[1] https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/bsds/
[2] http://scenebackgroundmodeling.net/
Any questions or comments, please contact Jingjing Zheng (jjzheng233@gmail.com) or Wenzhe Wang (woden3702@gmail.com).