Image
Segmentation.
What we have learn so far
Objective of Segmentation.
 Partitioning an image into region.
Early Approaches
 Found boundaries between regions based on discontinuities in intensity levels
Pixel Properties
 Accomplish segmentation via thresholds based on the distribution of pixel
properties.
In this section
 Segmentation techniques to find regions directly. .
2
1.
Region Based
Segmentation.
Region Growing
Region growing is a procedure that groups pixels
or sub regions into large region based on
predefine criteria for growth.
 Determine the threshold value
 Generate Seed Collection (close to the higher value of the predefine
properties)
 Calculate the pixel differences and compare it with Threshold value.
 If differences < or = threshold value mark it as region. (Concern the 8-
connected)
4
Example
5
6
Original Image
7
Seed Point : 255
8
Threshold: 255 - 255
9
Threshold : 190 - 255
10
Threshold : 155 - 255
11
12
13
Pros
Better in noisy image where
edges are hard to identify
Pros and Cons of Region Growing
Cons
Results depends on the
selected seed point.
14
Region Splitting and Merging
An alternative method for region image.
 Take full image and check overall pixels are homogeneous or not.
 If not divide image into 4 basic regions.
 Then check each regions are homogeneous or not and if not
divide again that region into 4 sub regions. Continue this process
until met homogeneous regions.
 Then merge if that adjacent separated regions have similar
properties.
15
Example
16
17
Original Image
18
Splitting Image
19
Merge and segmented image
Segmentation Using Morphological
Watersheds
 Based on visualizing an image in three dimension.
(spatial coordinates vs intensity : topographic
interpretation)
 Produce more stable segmentation results.
 Name refers to behavior of the geological watershed
which separated adjacent drainage basins.
20
How create topographic surface?
 High intensity denotes peeks and hills.
 Low intensity denotes valleys.
21
22
Use of Marker
 This watershed algorithm generally leads to over
segmentation due to noise and other local
irregularities of the gradient.
 Concept of marker is good approach to control over
segmentation.
 Internal marker : object of interest
 External marker : background
23
24
The use of motion in segmentation
 Used by humans and many other animals to extract
objects or regions of interest from a background of
irrelevant detail.
 Used in Robotic applications
Autonomous navigation
Dynamic scene analysis
25
Spatial Techniques
 Simple approaches for detecting changes between
two image frames (Compare two images pixel by
pixel)
26
Frequency Domain Techniques
 Compute Fourier Transformation of the image
Question No 01 :
What is the most stable segmentation method?
Morphological Watersheds Segmentation.
27
Question No 2 :
What are the pros and cons of region growing?
Pros
Better in noisy image where edges are hard to identify
Cons
Results depends on the selected seed point.
28
Question No 03:
What is the purpose of use markers in watershed
segmentation?
Prevent Over Segmentation.
29
30
Thank you!

Image segmentation

  • 1.
  • 2.
    What we havelearn so far Objective of Segmentation.  Partitioning an image into region. Early Approaches  Found boundaries between regions based on discontinuities in intensity levels Pixel Properties  Accomplish segmentation via thresholds based on the distribution of pixel properties. In this section  Segmentation techniques to find regions directly. . 2
  • 3.
  • 4.
    Region Growing Region growingis a procedure that groups pixels or sub regions into large region based on predefine criteria for growth.  Determine the threshold value  Generate Seed Collection (close to the higher value of the predefine properties)  Calculate the pixel differences and compare it with Threshold value.  If differences < or = threshold value mark it as region. (Concern the 8- connected) 4
  • 5.
  • 6.
  • 7.
  • 8.
  • 9.
  • 10.
  • 11.
  • 12.
  • 13.
  • 14.
    Pros Better in noisyimage where edges are hard to identify Pros and Cons of Region Growing Cons Results depends on the selected seed point. 14
  • 15.
    Region Splitting andMerging An alternative method for region image.  Take full image and check overall pixels are homogeneous or not.  If not divide image into 4 basic regions.  Then check each regions are homogeneous or not and if not divide again that region into 4 sub regions. Continue this process until met homogeneous regions.  Then merge if that adjacent separated regions have similar properties. 15
  • 16.
  • 17.
  • 18.
  • 19.
  • 20.
    Segmentation Using Morphological Watersheds Based on visualizing an image in three dimension. (spatial coordinates vs intensity : topographic interpretation)  Produce more stable segmentation results.  Name refers to behavior of the geological watershed which separated adjacent drainage basins. 20
  • 21.
    How create topographicsurface?  High intensity denotes peeks and hills.  Low intensity denotes valleys. 21
  • 22.
  • 23.
    Use of Marker This watershed algorithm generally leads to over segmentation due to noise and other local irregularities of the gradient.  Concept of marker is good approach to control over segmentation.  Internal marker : object of interest  External marker : background 23
  • 24.
  • 25.
    The use ofmotion in segmentation  Used by humans and many other animals to extract objects or regions of interest from a background of irrelevant detail.  Used in Robotic applications Autonomous navigation Dynamic scene analysis 25
  • 26.
    Spatial Techniques  Simpleapproaches for detecting changes between two image frames (Compare two images pixel by pixel) 26 Frequency Domain Techniques  Compute Fourier Transformation of the image
  • 27.
    Question No 01: What is the most stable segmentation method? Morphological Watersheds Segmentation. 27
  • 28.
    Question No 2: What are the pros and cons of region growing? Pros Better in noisy image where edges are hard to identify Cons Results depends on the selected seed point. 28
  • 29.
    Question No 03: Whatis the purpose of use markers in watershed segmentation? Prevent Over Segmentation. 29
  • 30.