Interface to move objects in an image in a user-friendly way.

Basically, this program implements part of the method described in the article Object Repositioning Based on the Perspective in a Single Image by S.Iizuka, Y. Endo, M. Hirose, Y. Kanamori, J. Mitani, Y. Fukui. (http://www.npal.cs.tsukuba.ac.jp/~iizuka/projects/reposition/data/cgf_repos.pdf)

• cmake
• SDL2
• OpenCV (The project uses version 2.4.12.3-1. It may not work with OpenCV 3)

cmake .
make

Input images are available in the images folder

Syntax :

./object_repositioning image_to_load

For example :

./object_repositioning images/beach.jpg

1. Wait for a few seconds until the image is shown (superpixels computation is done at launch)

2. Click where you want to place the ground line (max depth in the scene)

3. Draw a bounding box around the object you want to move

4. Wait (~1min depending on your processing power). When computation is finished, only the segmented object is shown.

5. Use the keypad to go through all the visualization modes

0 - Interactive mode. You can drag&drop a copy of the object you chose in the scene. The size of the object changed in real-time according to where you place it

1 - Superpixels

2 - Intersection between superpixels and bounding box * White : superpixels outside the bounding box * Black : superpixels which intersect the bounding box * Grey : superpixels inside the bounding box

3 - Saliency map

4 - Background probabilities (the more white a pixel is, the more likely it is to belong to the background)

5 - Foreground probabilities (the more white a pixel is, the more likely it is to belong to the foreground)

6 - Smoothness term for graph cut (weight between a pixel and its right neighbor)

7 - Smoothness term for graph cut (weight between a pixel and its bottom neighbor)

8 - Segmentation result (extracted object)