TELKOMNIKA
Vol. 10, No. 3 540
.
where
here R, G and B represent RE
The algorithm for noise remov
Step 1
: Divide the input retina
Step 2 : Use histogram equaliz
Step 3
: Use a 3x3 median filte
Step 4 : Convert the equalized
Step 5 : Calculate
N noise fac
Step 6 : Select a threshold valu
Step 7
:
for
each pixel, Calculate
NI Thres
if true
, add pixel in no
if false , add pixel in
end_for
Figure 2 shows the images. These segmentation
removed in fine segmentation.
Figure 2. Coarse segmentatio
2.2 Fine Segmentation
Background and n contain single pixel noise and
from segmentation masks. In dilation, morphological erosio
noise from binary masks [19]. operations [19]. Background s
order to remove the black pix removes all black single p
segmentation mask. Noise seg Figure 2. In order to remov
opening followed by erosion. single pixel noise and it giv
segmentation masks and fine
3, September 2012 : 537 – 544 RED, GREEN and BLUE components of RGB retina
oval mask is as below: nal image
Ii,j into non-overlapping blocks with size
lization to enhance the contrast ilter to reduce the noise in background of image.
ed and filtered RGB retinal image into HSI color spa factor which is a ratio of Hue and Intensity
alue empirically. reshold?
normal retinal image area pixels in noisy area pixels
e coarse background and noise segmentation m on masks contain single pixel and edge pixel no
on.
tion. a Original retinal image; b Background segm Noise segmentation mask
noise segmentation masks that are formed by coa nd edge pixels. Fine segmentation is done to rem
In fine segmentation, morphological operations i ion and morphological opening are applied to rem
9]. We have used 5 x 5 square structuring element d segmentation mask contains black single pixel no
pixel noise, square structuring element is used for pixel noise and edge pixels and it gives a
segmentation mask contains white single pixel noise ove the white pixel noise, square structuring ele
n. Opening removes all edge pixels and erosion r gives a fine noise segmentation mask. Figure
e segmentation masks. ISSN: 1693-6930
3
4 tinal image.
5 6
ize w x w
. space.
masks for retinal noise which will be
mentation mask; c
oarse segmentation move these noises
s i.e. morphological remove single pixel
nt for morphological noise Figure 2. In
for dilation. Dilation a fine background
ise and edge pixels lement is used for
n removes all white 3 shows coarse
TELKOMNIKA
Retinal Image Figure 3. Fine segmentati
Segme 2.3 Final Segmentation Mask
Final segmentation m and fine noise segmentation
removed by filtering the comb segmentation mask is then a
final segmentation masks for r
Figure 4. a: Original Color from database; b: Fine
Segmentation Mask; c: Segmentation Mask; d: Fina
Mask
3. Experimental Results
In medical image pr systems are very important
diaretdb0 [12], diaretdb1 [20], method. Diaretdb0 and Diar
resolution of 1500 X 1152 pi DRIVE and STARE database
objective and subjective crit segmented retinal images and
Statistical results of o table 1 and table 2. Table 1 s
noise segmentation mask and segmentation, fine segmen
segmentation after applying m accurate segmented retinal im
evaluate the proposed system experts and these manually la
3 shows the accuracy res segmentation masks.
ISSN: 1693-6930
e Preprocessing: Background and Noise Segmenta ation. a Original Color Retinal Images from databas
entation Masks; c Fine Segmentation Mask