image-inpainting-app

A Qt-based application to restore images with various inpainting methods.

This is also a reservoir for image inpainting algorithms in C++ (for personal amusement).

Original	Inpainted

Installation

The application is fully packed inside a Docker container, check these instructions to get started.

The app has been tested on the following platforms:

• Pop!_OS 22.04 LTS with Qt 6.8.1, Qt Creator 15.0.0 and GCC 11.4.0.

Usage

The app consists of a self-explanatory GUI and it contains a useful Help file. The following table explains the functionality for the GUI icons:

Icon	Functionality
	Import image
	Save image
	Show information about the application
	Activate pencil drawing mode (only available in Noise image mode)
	Adjust pencil properties
Original image	Show original image
Noisy image	Show noisy image
Inpainted image	Show inpainted image
Inpainting mask	Show inpainting mask
Clear all	Clear all images and setup
Reset draw	Reset pencil drawing
Noise	Add noise to the image
XYZ Image Inpainting	Use your favourite image inpainting algorithm

Available algorithms

Algorithm	Description	Link / Reference
BurgersViscousImageInpainting	Implements the inpainting algorithm based on the viscous Burgers' equation	Burgers Equation
CahnHilliardImageInpainting	Implements the inpainting algorithm based on the Cahn-Hilliard equation	Cahn-Hilliard Equation
MaxwellHeavisideImageInpainting	Implements the inpainting algorithm based on the Maxwell-Heaviside equations	Maxwell-Heaviside Theory
NavierStokesImageInpainting	Executes the image inpainting algorithm from OpenCV based on Navier-Stokes equations	OpenCV Navier-Stokes
TeleaImageInpainting	Executes the TELEA algorithm from OpenCV	Telea Inpainting
CriminisiImageInpainting	Implements Criminisi inpainting, based on cheind	Criminisi
FastDigitalImageInpainting	Implements Fast Digital Image Inpainting, based on this	Paper
LaplacianImageInpainting	Implements Laplacian Digital Image Inpainting, based on this	Paper
HarmonicImageInpainting	Heat (Harmonic) Equation Image Inpainting	Carola-Bibiane Schönlieb
...	...	...

---

TODO

• Mejorar lo de esconder y mostrar las labels. Por ejemplo, unir en un solo el label de mensaje temporal y tipo de imagen.
• Crear GIF de funcionamiento para el README
• Meter estadísticas de PSNR, MSE y SSIM tras hacer el inpainting
• Explicaciones de app e iconos en README. Enlazar el README en hipervínculo desde el About de la aplicación
• Explicaciones someras con ejemplo visuales de qué algoritmo va mejor para qué caso
• Gran problema: cuando se cambia grosor del lápiz, que no afecte al dibujo anterior, o sea, guardar el histórico de tuplas paths-grosor, no solo path.
• No aparece al brush style por defecto en docker, al cambiar el estilo del lápiz, por qué.
• Distribuir la imagen docker de la aplicación en Docker Registry o subirla el .tar a algún lado
• Release the app in AppImage format
• Implementar los algoritmos comentados en este README
• Diseñar un botón de "parar algoritmo" para abortar el procesamiento de CalculationThread
• Añadir el parámetro drawWhites al tipo de ruido; no sé qué problema daba el validador con tipo booleano.

References

• Inspired by GuidedDenoising