3-D discrete shearlet transform

Table of Contents

0. Introduction
1. Citation
2. Disclaimer and known issues
3. Installation
4. Gaussina Denoising Demo

Introduction

This repository contains the MATLAB/Octave implementation of 3D discrete shearlet transform and a video denosing demo using 3Dshearlet

By Pooran Singh Negi and Demetrio Labate

Department of Mathematics, University of Houston

Disclaimer

This software is provided "as-is", without any express or implied warranty. In no event will the authors be held liable for any damages arising from the use of this software.

Citation

If you use 3D Shearlet in your research, please cite:

@article{negilabate20123,
     title = {3-D discrete shearlet transform and video processing},
     author = {Negi, Pooran Singh and Labate, Demetrio},
     journal = {IEEE transactions on Image Processing},
     volume = {21},
     number = {6},
     pages = {2944--2954},
     year = {2012},
     publisher = {IEEE}
}

INSTALLATION

Ubuntu

Using octave

See this link for instllation instruction of octave in ubuntu Linux

https://askubuntu.com/questions/645600/how-to-install-octave-4-0-0-in-ubuntu-14-04

Change convnfft in shDec to convn if you plan to use octave to run this code. Code can run slow

MATLAB

For running 3DShearlet code, user need to run convnfft_install.m from CONVNFFT_Folder first.

Please use latest CONVNFFT from https://www.mathworks.com/matlabcentral/fileexchange/24504-fft-based-convolution to make sure it is compatible with latest MATLAB release.

Restriction on the size of .mat file

Due to nature of upsampling and downsampling for better denoising performance dimension of data need to be divisible by 3*2^(number of decomposition level required).

DEMO

Gaussian Denoising script is in DenoiseDemo folder with file name main.m

If user need to save data please use appropriate save statement after deonising is done. Current script setting denoises coastguard_sequence with simulated noise. This script can be run in background in Unix/Linux using hmatbg in DenoiseDemo folder. ./hmatbg main.m outfile hmatbg can be modified to change priority level using nice command.

For data corrupted by unknown Gaussian noise parameter, sdest3 function in Util directory can used to estimate standard deviation of the noise which is based on median of the wavelet coefficients at the finer scale.

NOTE: Files in 3DBP directory for doing bandpass are from SurfBox toolbox implemented by Yue Lu and Minh N. Do