Automatic seeded region growing for color image segmentation

advertisement
Automatic seeded region growing
for color image segmentation
Authors: Frank Y. Shih, Shouxian Cheng
Source: Image and Vision Computing,
vol. 23, pp.877-886, 2005.
Speaker: Shu-Fen Chiou(邱淑芬)
Date: 4/8/2015
1
Outline




Introduction
Proposed method
Experimental results
Conclusion
2
Introduction

Color image segmentation polices:




Threshold
Boundary-based
Region-based
Hybrid techniques
3
Hybrid techniques

Seeding region growing (SRG)

Different merging order possibility
SRG
4
Proposed method
1.
2.
3.
4.
Transform the color image from
RGB to YCbCr
Automatic seed selection algorithm
Region growing
Region merging
5
Transform the color image
from RGB to YCbCr

Why ?




YCbCr color space is widely used in video
compression standards. (e.g. MPEG and JPEG)
The color difference of human perception can be
directly expressed by Euclidean distance in the color
space.
The intensity and chromatic components can be
easily and independently controlled.
How ?
Y   65.481 128.553 24.966   R  16 
C   - 39.797 - 74.203 112   G   128
 b 
    
Cr   112
- 93.786 - 18.214  B  128
The Cb (Cr, respectively)is the difference between the blue (red, respectively)
6
component and a reference value
Automatic seed selection algorithm

Condition 1:


A seed pixel candidate must have the
similarity higher than a threshold
values.
Condition 2:

A seed pixel candidate must have the
maximum relative Euclidean distance
to its eight neighbors less than a
threshold value.
7
Automatic seed selection algorithm

The similarity of a pixel to its neighbors
is defined as : H  1   N




N



Y
  C b   Cr

 max
Considering 3 3 neighborhood, the
standard deviations :   1 x  x 2

x
9
High similarity
Condition 1:
Otsu’s method
A seed pixel candidate
must have the similarity
higher than a threshold
values.
8
Automatic seed selection algorithm

Calculate the maximum distance to its
8
neighbors as : d
 max d
max




di 
i
i 1
Y  Yi 2  Cb  Cb 2  Cr  Cr
i

2
i
, i  1...8
Y 2  Cb  Cr
YCbCr of the pixel, Yi Cbi Cri of its neighbors
2
Not on the boundary
0.05
2
Condition 2:
A seed pixel candidate
must have the maximum
relative Euclidean
distance to its eight
9
neighbors less than a
Automatic seed selection algorithm

Connected seeds are considered as
one seed.
Original color image
the detected seeds are shown in red color
10
Region growing

The pixels that are unclassified and
neighbors of at least one region,
calculate the distance:
Y  Y   C
2
di 
i
bi
 
2
 Cb  Cr i  Cr
Yi  Cbi  Cri
2
2

2
2
11
Region growing
1. red pixels are the seeds and the green pixels are the pixels
in the sorted list T in a decreasing order of distances.
2. the white pixel is the pixel with the minimum distance to
the seed regions
3. check its 4-neighbors
12
Region growing
1. If all labeled neighbors of p have a same label, set p to this label.
2. If the labeled neighbors of p have different labels, calculate the
distances between p and all neighboring regions and classify p to the
nearest region.
3. Then update the mean of this region, and add 4 neighbors of p, which
are neither classified yet nor in T, to T in a decreasing order of
distances.
4. Until the T is empty.
13
Region merging

Consider the color different and size of
regions :
Color different between two adjacent region
Ri and Rj is defined as:
2
2
2
Yi  Y j  Cbi  Cb j  Cri  Cr j
d Ri, Rj 
2
2
2
2
2
2 

min Yi  Cbi  Cri , Y j  Cb j  Cri 


 Size
1



 
 

select 150 of the total number of pixels in an
image as the threshold.
14
Region merging

We first examine the two regions
having the smallest color different
among others.



If d Ri, Rj < threshold, merge the two
regions and re-compute the mean of
the new region.
We repeat the process until no region
has the distance less than the
threshold.
Threshold=0.1
15
Region merging

If the size with number of pixels in
a region is smaller than a threshold,
the region is merged into its
neighboring region with the
smallest color difference.

This procedure is repeated until no
region has size less than the threshold.
16
Experimental results
JSEG algorithm.
17
Experimental results
JSEG algorithm.
JSEG algorithm.
18
Conclusion


We have presented an efficient
segmentation algorithm for color
image with automatic seed selection.
Experimental results show that we
have a better results than JSPG.
19
Download