DVDeconv

Deconvolution Matlab toolbox for fluorescence microscopy.

DVDeconv contains below 3 folders.

• PSFGenerator_GUI
• Deconv_GUI
• DataSet

"PSF Generator" and "Deconv_GUI" include Matlab GUI based tools for generating PSF and restoring images, respectively.
"DataSet" includes a script for generating dataset.

Usage

PSFGenerator_GUI

1. Launch Matlab.
2. Add path.
   addpath('PSFGenerator_GUI')
3. Run.
   PSFGenerator
4. After inputting parameters, click 'Generate' button.

Deconv_GUI

1. Launch Matlab.
2. Add path.
   addpath('Deconv_GUI')
3. Run.
   Deconv_GUI
4. Click "Open" in Image section. And select folder that include images to restore.
   (ex) DataSet/G10-P
5. Click "Open" in PSF section. And select folder that include PSF.
   (ex) DataSet/variantPSF
6. Select algorithm and setting parameters in Algorithm section.
   (ex) GEM algorithm, γ = 0.0005 , c = 5
7. After setting options for deconvolution in Result section, click "Run".

DataSet

When you want to generate bars image under depth-variant image model, refer dataset_variant.m script.
When you want to generate bars image under depth-invariant image model, refer dataset_invariant.m script.

Estimating aberration values

(main function) DVDeconv/PSFGenerator_GUI/ToFindParameters.m
(loss function) DVDeconv/PSFGenerator_GUI/myobjfun.m
When PSF is unknown, aberration values for PSF generation are estimated by referring to the code above.

Demo in the paper

PSFGenerator_GUI

1. Launch Matlab.
2. Add path.
   addpath('PSFGenerator_GUI')
3. Run.
   4-1. Leaving the initial values, click 'Generate' button.
   Then, the PSFs for depth-variant asymmetric algorithms are generated and saved in the current folder.
   The PSF for depth-invariant asymmetric algorithm is h00064.mat
   The generated results can be downloaded from 'DataSet/DVAsymPSF', 'DataSet/InvAsymPSF'
   4-2. Set W8 and W12 to 0, click 'Generate' button.
   Then, the PSFs for depth-variant symmetric algorithms are generated and saved in the current folder.
   The PSF for depth-invariant symmetric algorithm is h00064.mat
   The generated results can be downloaded from 'DataSet/DVSymPSF', 'DataSet/InvSymPSF'

Deconv_GUI

1. Launch Matlab.
2. Add path.
   addpath('Deconv_GUI')
3. Run.
   Deconv_GUI
4. Click "Open" in Image section. And select folder that include images to restore.
   For 15dB Dataset, set the path as 'DataSet/G15-P'
   For 10dB Dataset, set the path as 'DataSet/G10-P'
5. Click "Open" in PSF section. And select folder that include PSF.
   Select the PSF folder that matches the algorithm you want to execute.
   ex) DataSet/DVAsymPSF for Depth-variant asymmetric OSL/GEM algorithm
   ex) DataSet/InvSymPSF for Depth-invariant symmetric OSL/GEM algorithm
6. Select algorithm and setting parameters in Algorithm section.
   In the experiment in the paper,
   γ was set to 0.00001, c was set to 0.6 in the 15dB dataset.
   γ was set to 0.0006, c was set to 5 for depth-variant GEM algorithm in the 10dB dataset.
   γ was set to 0.0001 for depth-variant algorithm except GEM in the 10dB dataset.
7. After setting options for deconvolution in Result section, click "Run".
   The Depth-variant deconvolution results are obtained after 29 iterations.
   The Depth-invariant deconvolution results are obtained after 20 iterations.

References and License

If you find my work helpful in your research, please kindly cite below papers.

References:

[1] Kim, B., & Naemura, T. (2015). Blind depth-variant deconvolution of 3D data in wide-field fluorescence microscopy. Scientific reports, 5, 9894.
[2] Kim, B., & Naemura, T. (2016). Blind deconvolution of 3D fluorescence microscopy using depth‐variant asymmetric PSF. Microscopy research and technique, 79(6), 480-494.
[3] Kim, B (2021). An Open-Source MATLAB Toolbox for Depth-Variant Asym-metric Deconvolution of Fluorescence Micrographs. (in submission)

If you have any questions please feel free to contact me at by1110.kim@samsung.com

Author: Boyoung Kim
Copyright (c) 2021 Boyoung Kim
All rights reserved.