This is an implementation of the Interactive Digital Photomontage system presented in Agarwala, Aseem, et al. "Interactive digital photomontage." ACM Transactions on Graphics (TOG). Vol. 23. No. 3. ACM, 2004.
This version implements only the DESIGNATED_SOURCE data penalty and COLORS,GRADIENTS and COLORS_AND_GRADIENTS interaction penalties. This allows the user to composite images by painting the zone he wants to keep. The images are patched together using graph-cut optimization.
This project was the final paired-project for the Computer Vision class (INF552) at Ecole polytechnique (X) taught by Pr. Renaud Keriven. The webpage for the class can be accessed here.
-
Clone the repository
-
cmake CMakeLists.txt && make -
./PROJET -
An openCV window continaing the first image appears, using the mouse and the left button, you can paint on the zone you want to keep from the first image. Right-click on the image when you're done in order to go to the following image.
-
Strokes have different colors depending on the image they were defined on, in order for the user to remember the previous zones he chose.
-
Right-clicking on the last image launches the graph-cut optimization.
-
When graph-cut optimization is done, the result image appears along with the pixel map that shows the origin of each pixel in the result composite.
-
Tap any key to exit the program.
Initial photos:
Result photos:
