Functions > Image Processing > Segmentation > Example: Region Growing
  
Example: Region Growing
The reg_grow function divides an image into several homogenous connected regions using a region-growing algorithm. Region-based segmentation is used to group regions in an image that bear homogeneous properties, such as intensity, texture, and so on.
For information on using this example, refer to About Image Processing Examples.
1. Create an image with several rectangular boxes:
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
2. Add zero-mean uniform-distributed noise within [-0.1 0.1]:
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
3. View the image:
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
(reg_grow_s.bmp)
4. Use the region-growing algorithm:
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
5. View the output in false color to make the regions more obvious.
Click to copy this expression
Click to copy this expression
(reg_grow_sm1.bmp)
6. View the output in false color to make the regions more obvious.
Click to copy this expression
Click to copy this expression
Click to copy this expression
(reg_grow_sm1c.bmp)
7. Verify the number of regions found by the algorithm and then take a look at the histogram:
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
As in the input matrix, there are five regions with area of 400, two regions of 800, and three regions of 2000.
8. Apply this segmentation to a real image, an MRI image of a person's head.
Click to copy this expression
Click to copy this expression
Click to copy this expression
9. Extract the first 256 rows from the image to avoid having an odd number of rows:
Click to copy this expression
Click to copy this expression
Click to copy this expression
Click to copy this expression
10. Apply the region growing procedure to this image starting with an initial 2 x 2 partition and ending with 20 regions:
Click to copy this expression
11. Display the original image next to the segmented and scaled image.
Click to copy this expression
Click to copy this expression
(brain_t.bmp)
(brain_t1s.bmp)
12. Select all points in the segmented image which have the same value as the selected spoint.
Click to copy this expression
Click to copy this expression
Click to copy this expression
T2 is a binary image:
(brain_t2.bmp)
13. Use T2 as a mask to extract the brain region out of the original image.
Click to copy this expression
Click to copy this expression
(brain_extract.bmp)