Image Segmentation

Image Segmentation Chapter 10 1

Image Segmentation Segmentation subdivides an image into its constituent regions or groups. The level to which the subdivision is carried depends on the problem being solved. That is, segmentation should stop when the objects of interest in an application have been isolated. e.g. automated inspection of electronic assemblies; specific anomalies; missing components or broken connection paths.

2

Image Segmentation Segmentation algorithms ; Two categories based on two basic properties of intensity values : discontinuity and similarity First Category : Abrupt changes in intensity ; edges Second Category : partitionning of regions which are similar according to a set of predefined criteria. e.g. Thresholding, region growing, region splitting

3

Image Segmentation

First Category : Points, Lines, Edges

4

Detection of discontinuities Points, lines, edges The most common way

R = w1*z1 + w2*z2 + ……+ w9*z9

5

Point detection

R ≥ T T = Threshold

Figure 10.2 (a) point detection mask

6

Point detection

Figure 10.2

(b) X-ray image (c) Result of of a turbine point detection blade with

(d) Result of point detection mask with

7

Line detection – A Suitable Mask in desired direction – Thresholding

Figure 10.3 Line masks

8

Line detection • Example: Figure 10.4 Illustration of line detection (a) ,(b),(c)

-45º Mask

Thresholding 9

Edge Detection – Two Mathematical model

10

Edge Detection Gray level profile

First derivative

Second derivative

11

Problem of Noise

Gaussian Noise (mean, sigma)

12

Gradient Operators

13

Gradient Operators

Y-direction

X-direction

14

Gradient Operators – Roberts Cross Gradients:

– Prewitt Operators:

15

Diagonal Edge

45-Direction

– 45Direction

16

Gradient Operators

17

Gradient Operators

PreSmoothing 5×5

18

Diagonal edge detection

19

Laplacian as an isotropic Detector:

Discrete Implementation:

20

Laplacian of Gaussian (LoG):

21

Edge detection (overview)

22

Image Segmentation

Second Category : Thresholding, region growing, region splitting and merging

23

Thresholding – F(x,y)>T then (x,y) is belong to object, else (x,y) is belong to background. • Bi-level (T) • Multi-level (T1,T2,…, Tn) • Threshold image:

– Threshold Estimation : • Histogram

24

Thresholding

25