Computer vision

Dr.Eng. Mahmoud Abu_alfutuh

 Chapter 4
Segmentation
This chapter includes:
• Introduction
• Local Segmentation
• Global Segmentation
Machine vision system and Segmentation
• Segmentation is the initial stage for image
analysis, for recognition process.
 Question: True (T) & False (F).
Circle one
1 Feature Extraction is the next step in machine vision system after T F
Segmentation
2 Segmentation is the next step in machine vision system after Image T F
Acquisition
3 We can extract the edge in the image Y using the command: T F
edge(‘ canny ', Y);
4 Segmentation algorithms are based on discontinuity or similarity. T F

5 Global Segmentation seeks to detect edges within an image. T F

6 Sobel Operator uses First Order Derivative for edge detection. T F
7 We can extract the edge in the image i using the command: T F
edge(i,‘ prewitt ');
 Multiple choice questions
Circle one
1 Example of similarity approach in image segmentation is
  a Edge based c Boundary based segmentation
segmentation
  b Region based d Both a and c
segmentation
2 ______ refers to the process of partitioning an image into multiple
objects or parts.
  a Feature Extraction c Classification
  b Preprocessing d Segmentation
 Multiple choice questions
Circle one
1 Response of Sobel mask is zero at
  a Sharp intensities c Constant intensities
  b Low intensities d High intensities

2 Matlab code to extract the edge in the image i is

  a w = edge(i,‘ gaussian '); c w = edge( ‘ gaussian ', i);
  b w = edge(i,‘ sobel '); d w = edge( ‘ sobel ', i);
 Multiple choice questions
Circle one
Consider the following Image (F) which has integer intensities in the
range between 0 and 15, and then answer the following questions:

  4 3 2 2 4 2 2 3 4 2
Image F 2 4 3 4 3 2 3 4 2 2
3 2 3 4 4 4 2 2 4 3
1 Use Prewitt masks to compute the gradient (gx, gy) at the pixel F (2, 3) of
the image F.
  a gx= 2, gy= 1 c gx= 1, gy= 0
  b gx= 0, gy= 1 d gx= 1, gy= 1
2 Use Sobel masks to compute the gradient (gx, gy) at the pixel F (2, 3) of
the image F.
  a gx= 2, gy= 1 c gx= 3, gy= 1
  b gx= 1, gy= 3 d gx= 3, gy= 1
Sobel operators
the gradient at the pixel F (2, 3) :
g x = -1x3 - 2x2 -1x2 −𝟏
  −𝟐 −𝟏
+0x4 + 0x3 + 0x4 𝟎 𝟎 𝟎
+1x2+ 2x3 + 1x4 = 3 𝟏 𝟐 𝟏

 
−𝟏 𝟎 𝟏
g y = -1x3 + 0x2 +1x2
-2x4 + 0x3 + 2x4 −𝟐 𝟎 𝟐
-1x2 + 0x3 + 1x4 = 1
the gradient at the pixel F (2, 3) : −𝟏 𝟎 𝟏
 
𝑔𝑥 3
[ ] []
𝑔𝑦
=
1
 Multiple choice questions
Circle one
Consider the following Image (F) which has integer intensities in the
range between 0 and 15, and then answer the following questions:
  4 3 2 2 4 2 2 3 4 2
Image F 2 4 3 4 3 2 3 4 2 2
3 2 3 4 4 4 2 2 4 3
1 Find the size of the image (F) in bytes
  a 15 c 30
  b 3.75 d 11.25
2 Find the intensity of the pixels F (2, 5) in the above image when using 3×3
Average filter
  a 1.93 c 19
  b 3.22 d 1.9
 Multiple choice questions
Circle one
Consider the following Image (F) which has integer intensities in the
range between 0 and 15, and then answer the following questions:
  4 3 2 2 4 2 2 3 4 2
Image F 2 4 3 4 3 2 3 4 2 2
3 2 3 4 4 4 2 2 4 3
1 Find the histogram function at gray levels: 0, 2, 4. H(0), H(2), H(4)
  a H(0)=0, H(2)=11, H(4)=9 c H(0)=0, H(2)=10, H(4)=10
  b H(0)=0, H(2)=12, H(4)=10 d H(0)=0, H(2)=13, H(4)=10
2 Do histogram equalization on the image F, and then find the histogram
function at gray levels: 6, 10, and 15. H(6), H(10), H(15)
  a H(6)=8, H(10)=12, H(15)=10 c H(6)=12, H(10)=8, H(15)=10
  b H(6)=0, H(10)=12, H(15)=8 d H(6)=0, H(10)=12, H(15)=10
 histogram equalization
• To equalize this histogram, we form running totals of the
ni, and multiply each by (L-1)/n = 15/30 =1/2:
• Grey level i 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
• ni 0 0 12 8 10 0
• total 0 0 12 20 30 30 30 30 30
• total × o.5 0 0 6 10 15 15 15 15
 Multiple choice questions
Consider the following Image F which has integer intensities in the
range between 0 and 15, and then answer the following questions:
12 12 11 9 10 11 12 12
12 4 4 4 4 4 4 12
12 4 4 4 4 4 4 12
12 12 7 5 5 6 12 12

1 Find the two peeks of the histogram function of Image F and the
corresponding gray levels intensities
  a H(5)=12, H(12)=12 c H(7)=12, H(12)=12
  b H(4)=12, H(12)=12 d H(11)=12, H(12)=12
2 Find a Threshold value (T) to segment the image F
  a T=4 c T=8
  b T=5 d T=12
 histogram
• Grey level i 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
• ni 0 0 0 0 12 2 1 1 1 1 2 12 0 0 0
Define image segmentation 
image segmentation is the process of partitioning a
digital image into multiple segments (objects).
Image segmentation is used to locate objects and
boundaries in images.
The result of image segmentation is a set of
segments, or a set of contours extracted from the
image.
Discus briefly main approaches used in Segmentation
Segmentation approaches are:
- Pixel-based or local method
- Region-based or global approach.
1. The pixel-based approach seeks to detect edges within an
image and then link them to create a boundary which
encloses a region of uniformity.
2. The region-based approach seeks to create regions
directly by grouping together pixels which share common
features into areas or regions of uniformity.
List the methods of edge detection
• First Order Derivative / Gradient Methods
– Sobel Operator
– Prewitt Operator
• Second Order Derivative
– Laplacian
– Laplacian of Gaussian
• Optimal Edge Detection
– Canny Edge DetectionM

17
 Matlab code to find edges
• Write a Matlab code to find edges in grayscale image (circuit.tif)
using Canny operators. Shows the respective results.

• Matlab code
• >> ic=imread('circuit.tif ');
• >> c=edge(ic,‘ Canny ');
• >> subplot(1,2,1), imshow(ic), title('Original Image');
• >> subplot(1,2,2), imshow(c), title('Image after using Canny edge
detector')
List Region-based methods
Region-based methods are:
1. Image Thresholding
2. Region Merging and Splitting
3. Region Growing
Describe an algorithm for double thresholding a grayscale image. Write a
Matlab code to double threshold an image named (cell.tif), using threshold =
110 and 130. Shows the respective results.
•Algorithm for double thresholding a grayscale image
•For each pixel in the image do
• If (pixel_value > t1 and pixel_value < t2)
• then
• label the pixel as an object pixel;
• else
• label the pixel as a background pixel;
 
•Matlab code
•>> I=imread('cell.tif');
•>> imshow(I), figure, imshow(I>110 & I<130)
Write a 'split and merge algorithm' for segmenting an
image.
split and merge algorithm:
• It is a top-down approach and it starts with the assumption that
the entire image is homogeneous, If this is not true, the image is
split into four sub images
• If a region R is inhomogeneous, then is split into four sub regions
• If two adjacent regions Ri, Rj are homogeneous, they are merged
• The algorithm stops when no further splitting or merging is
possible
Thank you for listening

22