BLOTLESS: a new dictionary learning algorithm with BLOTLESS update. (Compared to the benchmark algorithms, BLOTLESS method works better when the number of training samples is small. This could be very useful in some cases that only a small dataset is given and we need to train a reasonable dictionary for sparse representation.)
=======================================================================
For more technical details please refer to our paper: Yu, Qi, Wei Dai, Zoran Cvetkovi?, and Jubo Zhu. "Dictionary Learning with BLOTLESS Update." IEEE Transactions on Signal Processing 68 (2020): 1635-1645.
If you find any issue, please let me know via email (yuqi10@nudt.edu.cn). I would really appreciate. Thank you.
Release date: 01/07/2020
========================================================================
Table of content
- [utilties]
- [ksvdbox13](ksvd toolbox)
- [ompbox10](omp toolbox)
- some other functions needed in the BLOTLESS algorithm
- [Blotless.m](A demo of BLOTLESS method using synthetic data.)
- [Blotless_learnDict.m](Learning a complete/over-complete dictionary from image samples data 'tdata4.mat')
- [Deoise_compare_allmethod](Denoising demo of images in our paper.)
- [tdata4.mat](Image patches from dataset: Olivetti Research Laboratory (ORL) face database, each column stands for a 8*8 patch)
- [3,5,6,7.pgm](Test face images used in our paper.)
========================================================================