You are on page 1of 4

IMAGE SEGMENTATION

Three approaches of segmentation:


1. Thresholding, i.e. distribution of pixels
properties
2. Discontinuity, i.e. to partition the image based
on abrupt changes in intensity (gray levels),
e.g. edges
3. Similarity, i.e. to partition the image into
SEGMENTATION BY SIMILARITY Dr. Nursuriati similar (according to predefined criteria)
regions, e.g. region growing, region splitting
Jamil
and merging

TOPICS REGION GROWING

1. Region Growing Start with a set of “seed” points and from


2. Split and Merge these grow regions by appending to each
seed point those neighboring pixels that
3. Watershed have similar properties.
Properties: gray level, color, texture, shape.

REGION GROWING REGION GROWING


Seed = 99
Steps and problems: Predicates:
Determine seed: depends on application.  Two pixels belong to the same region if |f(x0, y0) - f(x1, y1)|
Different choice of seeds give different <1
segmentations.  A pixel belongs to a region if it is N8 connected at least to
one pixel of the region.
Determine predicate(s). Selection of suitable
properties for including points in the various
regions.
Determine stopping rule.
Prior knowledge of the image important.

1
REGION GROWING REGION GROWING

Growing criteria:
Since weld defects pixels are very bright, good seed  Difference between pixel and seed less than 65.
pixel value is 255.  8-connected adjacency.

REGION GROWING
REGION GROWING
i=imread('pinktumor2.jpg');
f=rgb2gray(i);
[g,SI,TI]=regiongrow(f,75,70);
figure, subplot(121); imshow(f);
subplot(122); imshow(g);
Seeds

I = im2double(imread('medtest.png'));
x=198; y=359;
J = regiongrowing(I,x,y,0.2);
figure, subplot(131); imshow(I);
subplot(132); imshow(J);

Region growing subplot(133); imshow(I+J);


Segmented regions

REGION SPLIT & MERGE


• Subdivide an image initially into a set of
arbitrary, disjointed regions and then merge
and/or split the regions in an attempt to satisfy the
conditions of region-oriented segmentation.

• Quadtree-based algorithm

SEGMENTATION BY SIMILARITY
(SPLIT AND MERGE)
11

2
REGION SPLIT & MERGE STEP 1: SPLIT REGION R WHERE P(RI) =FALSE

Steps:
1. Split into 4 disjointed quadrants any region Ri where P(Ri) =
FALSE
2. Merge any adjacent regions Rj and Rk for which P(Rj ∪ Rk) =
TRUE
3. Stop when no further splitting or merging is possible.

STEP 1: SPLIT REGION R WHERE P(RI) =FALSE STEP 1: SPLIT REGION R WHERE P(RI) =FALSE

STEP 2. Merge adjacent regions Rj and Rk which P(Rj ∪ Rk) STEP 2. Merge adjacent regions Rj and Rk which P(Rj ∪ Rk)
= TRUE = TRUE

3
STEP 2. Merge adjacent regions Rj and Rk which P(Rj
∪ Rk) = TRUE

You might also like