ashkan-abbasi66 / nwsr Goto Github PK

Description:

This repository contains the code associated with the following paper:

Ashkan Abbasi, Amirhassan Monadjemi, Leyuan Fang, Hossein Rabbani, "Optical coherence tomography retinal image reconstruction via nonlocal weighted sparse representation," J. Biomed. Opt. 23(3), 036011 (2018), doi:10.1117/1.JBO.23.3.036011.

We introduce a method for OCT retinal image denoising and interpolation. If you find it useful, kindly cite the above paper.

Requirements:

Before running the code, you need to do the following steps:

1. Download datasets for training and testing from http://people.duke.edu/~sf59/Fang_TMI_2013.htm

2. Create a "datasets" folder and copy the followings two folders in it:

   For synthetic experiments

   Images for Dictionaries and Mapping leraning

3. Install SPAMS from http://spams-devel.gforge.inria.fr/downloads.html we called "mexOMP" function from SPAMS toolbox in the "sparse_inp_momp.m"

   Note: it is also possible to use "omp" function which is in the "OMPBox v10" toolbox. however, it needs small modifications in our "sparse_inp_momp.m". The "OMPBox v10" can be downloaded from: http://www.cs.technion.ac.il/~ronrubin/software.html

4. If you want to train a dictionary, you need to install "KSVDBox v13" which can be downloaded from http://www.cs.technion.ac.il/~ronrubin/software.html

Important m files:

demo_NWSR_1xInterp.m: Denoising experiment

demo_NWSR_2xInterp.m: Interpolation experiment

load_train_images.m: training High-SNR High-resolution (HH) images

noise_level_in_HH_training_images.m: estimate noise standard deviation in the training set

demo_train_dictionary.m: learn a dictionary from training HH images or show the atoms of a saved dictionary.

Note: the file "dictionary_8x8_20it_rand_g165.mat" contains the dictionary that we have trained with this function, as we explained in the paper.