of 14
8/20/2019 07-regionsegmentation
1/32
Region Segmentation
Computação Visual e Multimédia 10504: Mestrado em Engenharia Informática
Chap. 7 — Region Segmentation
8/20/2019 07-regionsegmentation
2/32
Chapter 7: Region Segmentation
Outline
8/20/2019 07-regionsegmentation
3/32
Chapter 7: Region Segmentation Image segmentation: a reminder
8/20/2019 07-regionsegmentation
4/32
Chapter 7: Region Segmentation Image segmentation: … in pictures
M e a n S h i f t : A R o b u s t A p p r o a c h t o w a r d F e a t u r e S p a c e A n a l y
s i s , b y D . C o m a n i c i u a n d P . M e e r
h t t p : / / w w w . c a i p . r u t g e r s . e d u
/ ~ c o m a n i c i / M S P A M I / m s P a m i R e s
u l t s . h t m l
Edge segmentation:
Region segmentation:
h t t p : / / r o b o t s . s t a n f o r
d . e d u / c s 2 2 3 b / i n d e x . h t m l
8/20/2019 07-regionsegmentation
5/32
Chapter 7: Region Segmentation
8/20/2019 07-regionsegmentation
6/32
Chapter 7: Region Segmentation
What is a region?
8/20/2019 07-regionsegmentation
7/32
Chapter 7: Region Segmentation
Region-based approach
8/20/2019 07-regionsegmentation
8/32
Chapter 7: Region Segmentation
Region-based segmentation
Ri= I
i=1
n
!
Ri R
j ="! #i =1,2,…n
P Ri
R j !( ) = FALSE
P( Ri)=
TRUE , "i
8/20/2019 07-regionsegmentation
9/32
Chapter 7: Region Segmentation Comparison of histogram, region growingand deformable contour segmentations
3 5 7 3 4 2 1
2
4
9
10
22
9
3
3 5 12 11 15 10 3
5
6
11
9
17
19
1
2 3 11 12 18 16 2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
3 5 7 3 4 2 1
2
4
9
10
22
9
3
3 5 12 11 15 10 3
5
6
11
9
17
19
1
2 3 11 12 18 16 2
3 6 8 10 18 9 5
4
6
7
8
3
3
1
3 5 7 3 4 2 1
2
4
9
10
22
9
3
3 5 12 11 15 10 3
5
6
11
9
17
19
1
2 3 11 12 18 16 2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
3 5 7 3 4 2 1
2
4
9
10
22
9
3
3 5 12 11 15 10 3
5
6
11
9
17
19
1
2 3 11 12 18 16 2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
3
5
7
3
4
2
1
2 4 9 10 22 9 3
3
5
12
11
15
10
3
5
6
11
9
17
19
1
2
3
11
12
18
16
2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
3
5
7
3
4
2
1
2 4 9 10 22 9 3
3
5
12
11
15
10
3
5
6
11
9
17
19
1
2
3
11
12
18
16
2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
3
5
7
3
4
2
1
2 4 9 10 22 9 3
3
5
12
11
15
10
3
5 6 11 9 17 19 1
2 3 11 12 18 16 2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
3
5
7
3
4
2
1
2 4 9 10 22 9 3
3
5
12
11
15
10
3
5
6
11
9
17
19
1
2
3
11
12
18
16
2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
3 5 7 3 4 2 1
2 4 9 10 22 9 3
3
5
12
11
15
10
3
5 6 11 9 17 19 1
2 3 11 12 18 16 2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
3 5 7 3 4 2 1
2
4
9
10
22
9
3
3
5
12
11
15
10
3
5 6 11 9 17 19 1
2
3
11
12
18
16
2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
3 5 7 3 4 2 1
2
4
9
10
22
9
3
3
5
12
11
15
10
3
5 6 11 9 17 19 1
2
3
11
12
18
16
2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
3 5 7 3 4 2 1
2
4
9
10
22
9
3
3
5
12
11
15
10
3
5 6 11 9 17 19 1
2
3
11
12
18
16
2
3 6 8 10 18 9 5
4 6 7 8 3 3 1
threshold T ≥10 threshold T !11 threshold T !12
region growing with variance of 2 in respect to value 11 with reference to threshold T !11
deformable contour to meet threshold T !11
8/20/2019 07-regionsegmentation
10/32
Chapter 7: Region Segmentation
Region growing segmentation
seed growing final region
h t t p : / / u e i . e n s t a . f r
/ b a i l l i e
8/20/2019 07-regionsegmentation
11/32
Chapter 7: Region Segmentation Seed-based region growing segmentation: pixel aggregation
The seed point can be selected either by a human or automatically byavoiding areas of high contrast (large gradient) => seed-based method.
8/20/2019 07-regionsegmentation
12/32
Chapter 7: Region Segmentation Seed-based region growingsegmentation: example
original image
threshold:225~255
threshold: 190~225
threshold = 255 returns multipleseeds
threshold: 155~255
h t t p : / / e n . w
i k i p e d
i a . o r g / w i k i / R e g i o n_ g r o w i n g
Problem: To isolate the strongest lightning region of theimage on the right hand side without splitting it apart. Solution: To choose the points having the highest gray-scale
value which is 255 as the seed points shown in the imageimmediately below.
8/20/2019 07-regionsegmentation
13/32
Chapter 7: Region Segmentation Fast scanning algorithm: a kind of unseeded region segmentation
1
Threshold T: P1 == P2 iff Diff (Col (P1),Col (P2)) < T
val=? y
x
x==y : val = x
xy :
boundary (x)!y if |x-y|≤T
new region index if |x-y|>T
1 2 2 3
1 1 1 1 1
If the criterion of homogeneity is local (compared to the value ofthe candidate pixel and the pixel of the border) => linear method.
h t t p : / / u e i . e n s t a . f r
/ b a i l l i e
8/20/2019 07-regionsegmentation
14/32
Chapter 7: Region Segmentation Region growing segmentation: advantages & disadvantages
h t t p : / / u e i . e n s t a . f r
/ b a i l l i e
8/20/2019 07-regionsegmentation
15/32
Chapter 7: Region Segmentation
Region splitting and merging segmentation
original image splitting & merging thresholding seg.
8/20/2019 07-regionsegmentation
16/32
Chapter 7: Region Segmentation Region splitting: example
original image split 1
split 2 split 3
In this example, the criterion of homogeneity is the variance of 1.
h t t p : / / u e i . e n s t a . f r
/ b a i l l i e
8/20/2019 07-regionsegmentation
17/32
Chapter 7: Region Segmentation
Splitting & merging: data structures
RAG with adjacency relations (inred ) for big black region.
8/20/2019 07-regionsegmentation
18/32
Chapter 7: Region Segmentation
Splitting & merging segmentation algorithm
RAG with adjacency relations (inred ) for big black region.
h t t p : / / a s t r o . t e m p l e . e d u / ~ s i d d u
8/20/2019 07-regionsegmentation
19/32
Chapter 7: Region Segmentation
Watershed segmentation
h t t p : / / e n . w
i k i p e d i a . o r g / w i k i / W a t e r s h e d_
( i m a g e_
p r o c e s s i n g ) # c i t e_
n o t e - 1
watersheds
watershed crest line
8/20/2019 07-regionsegmentation
20/32
Chapter 7: Region Segmentation Watershed segmentationby flooding
original image3D topographic surface
This technique aims at identifying all the third type ofpoints (i.e., points of watershed lines) for segmentation
8/20/2019 07-regionsegmentation
21/32
Chapter 7: Region Segmentation
Watershed segmentation by flooding
h t t p : / / e u c l i d . i i . m e t u . e d u . t r / ~ i o n 5 2 8 / d e m o
/ l e c t u r e s / 6 / 4 / i n d e x . h t m l
255
0
255
0
255
0
255
0
8/20/2019 07-regionsegmentation
22/32
Chapter 7: Region Segmentation
Watershed segmentation algorithm
8/20/2019 07-regionsegmentation
23/32
Chapter 7: Region Segmentation Watershed segmentation algorithm: dam construction
8/20/2019 07-regionsegmentation
24/32
Chapter 7: Region Segmentation Flooding-based watershed segmentationapplied to gradient image
original image gradient image
watershed of the gradient image
final contours
h t t p : / / c m m . e n s m p . f r / ~ b e u c h e r / w t s h e d . h t m l
8/20/2019 07-regionsegmentation
25/32
Chapter 7: Region Segmentation Marker-controlled watershedsegmentation ( gradient image )
original image over-segmented image
h t t p : / / c m m . e n s m p . f r / ~ b e u c h e r / w t s h e d . h t m l
markers of the blobsand of the background
marker-controlled watershedof the gradient image
8/20/2019 07-regionsegmentation
26/32
Chapter 7: Region Segmentation
Inter-pixel watershed segmentation
h t t p : / / e n . w
i k i p e d i a . o r g / w i k i / W a t e r s h e d_
( i m a g
e_ p r o c e s s i n g ) # c i t e_
n o t e - 2
8/20/2019 07-regionsegmentation
27/32
Chapter 7: Region Segmentation
More complex segmentation methods
8/20/2019 07-regionsegmentation
28/32
Chapter 7: Region Segmentation
Snakes
Copyright G.D. Hager Images taken from http://www.cs.bris.ac.uk/home/xie/content.htm
8/20/2019 07-regionsegmentation
29/32
Chapter 7: Region Segmentation
Level sets
Copyright G.D. Hager Images taken from http://www.cgl.uwaterloo.ca/~mmwasile/cs870/
8/20/2019 07-regionsegmentation
30/32
Chapter 7: Region Segmentation
Graph cuts
Copyright G.D. Hager Images taken from efficient graph-based segmentation paper
8/20/2019 07-regionsegmentation
31/32
Chapter 7: Region Segmentation Generalized PCA(Rene Vidal)
Copyright G.D. Hager
Human GPCA
8/20/2019 07-regionsegmentation
32/32
Chapter 7: Region Segmentation
Summary:
