Grow Cut is a simple and powerful segmentation technique. The algorithm takes two inputs: an intensity image and an initial set of seeds or strokes from the user depicting the foreground and background. The algorithm is iterative and needs to maintain two intermediate images: a strength image and the current labeling. Here, we pack the two intermediate images into a single texture using different color channels for each intermediate image.
The managment of multiple input textures is based on this work.
Select a color and then click and drag to paint input seeds. Lower image updates automatically.
| Red | Green | Blue | Clear |
Move pointer over the lower image. To the left of the mouse, the original intensity image is displayed with the pixels recolored to show the segmentation.
To the right of the mouse, the segmentation is shown above the mouse, while the strength image is shown below.
Check out the source code.
This demo uses WebGL. Not all devices and browsers are supported.