i. The segmentation which is carried out based on similarities in the given image is known as region based segmentation.

ii. The regions that are formed using this method have the following properties

a. The sum of all the regions is equal to the whole image.

b. Each region is contiguous and connected

c. A pixel belongs to a single region only, hence there is no overlap of pixels.

d. Each region must satisfy some uniformity condition

e. Two adjacent regions do not have anything in common.

iii. Region based segmentation can be carried out in four different ways :

(I) Region Growing

(II) Region Splitting

(III) Region merging

(IV) Split and merge

Each of them is explained below

(I) Region Growing

i. The procedure in which pixels are grouped into larger regions based on some predefined conditions is known as region growing

ii. The basic approach is to pick a seed point (Starting pixel) and grow regions from this seed pixel

iii. Let us pick up a random pixel (x1,y1) from the image that needs to be segmented. This pixel is called the seed pixel. The nearest neighbors of (x1,y1) are examined depending on the type of connectivity assumed (4 connectivity or 8 connectivity)

iv. The neighboring pixel is accepted in the same region as (x1,y1) if they together satisfy the homogeneity property of a region. That is, both of them satisfy a predefined condition of the region.

v. Once a new pixel say (x2,y2) is accepted as a member of the current region, the neighbors of this new pixel are examined to increase the region further.

vi. This procedure goes on until no new pixel is accepted. All the pixels of the current region are given a unique label. Now a new seed is chosen and the same procedure is repeated. This procedure is repeated till the time all the pixels are assigned to some group or the other.

(II) Region Splitting

i. In region splitting, we try to satisfy the homogeneity property where pixels that are similar are grouped together.

ii. If the grey levels present in the region do not satisfy the property, we divide the region into four equal quadrants. If the property is satisfied, we leave the region as it is.

iii. This is done repeatedly, until all the regions satisfy the given property.

iv. The splitting technique is shown below

v. In the above example, the entire image will be represented by R, this R is the parent node, it is split into four leaf nodes R1, R2, R3 and R4. Of these leaf nodes only R4 does not contain pixels which satisfy some common property, hence R4 is split into four regions R41, R42, R43 andR44.

vi. Now if all the pixels in a particular region are satisfying some common property in that region, then splitting is stopped. This is how region splitting works.

(III) Region Merging

i. The region merging method, is exactly opposite to the region splitting method

ii. In this method, we start from the pixel level and consider each of them as a homogeneous region. At any level of merging we check if four adjacent homogeneous regions arranged in a 2 x 2 manner, together satisfy the homogeneity property.

iii. If the property is satisfied, then the pixels are merged to form a bigger region, otherwise the regions are left as they are.

(IV) Split and Merge

i. Region splitting and region merging were explained above, in region splitting we start with the whole image and split the image into four quadrants. We continue splitting each quadrant further, until all the sub-regions satisfy the predefined homogeneity property.

ii. In Region merging each pixel is taken as a small region, we merge small regions into larger regions if they satisfy the homogeneity property.

iii. If the homogeneous regions are small, region merging technique is superior and if the regions are large, region splitting is preferred.

iv. Now, in most applications a combination of both, region splitting and region merging is used.

v. Such a technique where the above combination is used is known as Split and Merge technique.