Image restoration using PDEs

This is a simple project to test different image restoration methods, based on PDEs. The Image restoring method used here are simple diffusion and the Perona-Malik method. To run the programm main.py, this will launch a GUI in which you can choose the different options, if you prefer you can also run the script test.py and in this one change the parameters of image_name, rest_method and steps.

Libraries needed

This program was developed using Python3 and uses the libraries

• Numpy
• OpenCv2
• Scipy
• PyQt5
• Skimage

How to use the GUI

If you run the program main.py it will launch a GUI developed in PyQt5. The steps to follow are the next:

1. Pres Ctrl + L or click the button [...] to load a damaged image or open the menu file and select Load Image. The program only supporst .png, .bmp and .jpg formats for the moment.
2. Once you have choosen in the box Select a Method select a method to restore the image.
3. Select the number of steps on which the restoration method will be iterated.
4. If you have the original image before damaged and you want to select the Discrepancy Score, check the box Compare with original?
5. Click Go and wait for results.

IMPORTANT

• The methods for Image impainting, i.e. impaint_Diff, biharmonic_impainting and impaint_Perona_Malik only work if they have the original image to compare with, reason for which the option Compare with original?must be selected.
• The region selection is now added but it still have flaws with image of sizes of less than 300x300 pixels.
• Full size screen present some bugs

Example of Results

Image damaged

Imapinting using Perona-Malik for 600 steps

Imapinting using Perona-Malik for 3000 steps

Imapinting using Perona-Malik for 3000 steps