Image Processing Reference
In-Depth Information
FIGURE 3
Image binarization on crystallization trial images: (a) original images, (b) Green
percentile threshold (
p
= 95), and (c) Green percentile threshold (
p
= 99).
4.2 Region Features
After we generate the binary image, we apply connected component labeling to segment the
regions (crystals). The binary image can be obtained from any of the thresholding methods.
Let O be the set of the blobs in a binary image B, and B consists of
n
number of blobs. The
blobs are ordered from the largest to the smallest such that area (
O
i
) ≥ area(
O
i
+1
). Each blob
O
i
is enclosed by a minimum boundary rectangle (MBR) having width (
w
i
) and height (
h
i
). It
should be noted that the blobs may not necessarily represent crystals in an image. For such
cases, the blob features may not be particularly useful for the classifier.
Table 1
provides a list
of the region features. We extract five features (area, perimeter, filled area, convex area, and
eccentricity) from the three largest blobs
O
1
,
O
2
, and
O
3
. If the number of blobs is less than 3,
value 0 is assigned. We apply green percentile image binarization with
p
= 95 and
p
= 99. From
each binary image, we extract 17 region features. Thus, at the end of this step, we obtain 34
features.
Table 1
Region Features
Symbol
Term
Description
n
Number
of
Number of blobs with a minimum size of 25 pixels
blobs
ch
Convex
hull
Area of the convex hull (smallest set of pixels) that enclose all blobs
area
a
1
,
a
2
,
a
3
Blob area
Number of pixels in the three largest blobs
O
1
,
O
2
,
O
3
p
1
,
p
2
,
p
3
Blob perimeter
Perimeter of the three largest blobs
O
1
,
O
2
,
O
3
fa
1
,
fa
2
,
fa
3
Blob filled area
Number of white pixels in the three largest blobs
O
1
,
O
2
,
O
3
ch
1
,
ch
2
,
ch
3
Blob
convex
Number of pixels within the convex hull of the three largest blobs
O
1
,
O
2
,
O
3
area
e
1
,
e
2
,
e
3
Blob
eccentri-
Eccentricity of the three largest blobs
O
1
,
O
2
,
O
3
city
Search WWH ::
Custom Search