OPTIC DISC DETECTION IN RETINAL IMAGES BY PATTERN

DISTANCE MINIMIZATION

Marcy A. Dias and Fernando C. Monteiro

Polytechnic Institute of Braganc¸a, Campus Santa Apol´onia, Apartado 1134, 5301-857 Braganc¸a, Portugal

Keywords:

Earth Mover’s Distance, Gabor Wavelet Transform, Optic Disc Detection, Retinal Images.

Abstract:

The retinal fundus photograph is widely used in the diagnosis and treatment of various eye diseases such as

diabetic retinopathy and glaucoma. On the research work leading to automatic analysis of retinal images, the

knowledge of the optic disc (OD) location is essential, and a new method to locate the optic disc automatically

is proposed. We propose an algorithm for the detection of OD in the retina which takes advantage of the

powerful preprocessing techniques such as the contrast enhancement, Gabor wavelet transform, mathematical

morphology and Earth Mover’s distance as the matching process. Forty images of the retina from the DRIVE

database were used to evaluate the performance of the method.

1 INTRODUCTION

The retina is the inner-most layer of the eye where the

earliest pathological changes are seen. It is composed

of various anatomical structures which indicate many

diseases, such as hypertension, diabetic retinopathy

and glaucoma. Retinopathy is one of the main causes

of blindness in the working age population. Analysis

of retinal images is considered essential for diagnosis

and treatment of many diseases affecting the retina.

These images must be accurately segmented to extract

sensitive objects in the retina such as the blood vessel

tree, the optic disc, the macula and the region between

the optic disc and the macula.

Retinal or fundus images provide information

about the blood supply system to the retina. The op-

tic disc is a bright area within a retinal image and is

the exit point of retinal nerve ﬁbers from the eye and

the entrance and exit point for retinal blood vessels

(See Fig. 1(a)). Optic Disc detection is a reference

to locate the various anatomical features in the retinal

images.

Several OD detection techniques make use of

anatomical structures among the OD, macula, and

retinal blood vessels (Youssif et al., 2008). For exam-

ple, some methods are based on the anatomical struc-

ture that all major retinal blood vessels radiate from

the OD. Some other methods make use of the relative

position between the OD and the macula that often

varies within a small range. Compared with the im-

age characteristics, the anatomical structures are more

reliable under the presence of retinal lesion and imag-

ing artifacts. However, the extraction of either retinal

blood vessels or the macula is often a nontrivial task

by itself.

This paper presents a automatic method for optic

disc detection which takes advantage of the power-

ful preprocessing techniques such as the contrast en-

hancement, Gabor wavelet transform, mathematical

morphology and Earth Mover’s distance for matching

pattern. The methods include the design of a bank of

directionally sensitive Gabor ﬁlters for several values

of the scale and elongation parameters as proposed by

Soares et al. (Soares et al., 2006). Forty images of the

retina from the DRIVE database (Staal et al., 2004)

were used to evaluate the performance of the method.

The rest of this paper is organized as follows. Sec-

tion 2 describes the proposed OD detection technique.

Experimental results are then described and discussed

in Section 3. Some concluding remarks are ﬁnally

drawn in Section 4.

2 MATERIALS AND METHODS

The DRIVE database consists of 40 images (seven of

which present pathology), compressed in JPEG for-

mat of size 565 ×584 pixels, eight bits per colour

channel, obtained from a diabetic retinopathy screen-

ing program. The images are acquired using a Canon

CR5 nonmydriatic 3CCD camera at 45

ﬁeld of view.

356

A. Dias M. and C. Monteiro F..

OPTIC DISC DETECTION IN RETINAL IMAGES BY PATTERN DISTANCE MINIMIZATION.

DOI: 10.5220/0003766103560359

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS-2012), pages 356-359

ISBN: 978-989-8425-89-8

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

The 40 images were divided into two sets, a test set

and a training set, each containing 20 images. The

images have been manually segmented by three ob-

servers to be used as references for comparing the

computer-generated segmentations.

2.1 Contrast Adjustment

After converting each pixel in the image to a vec-

tor of colour components and normalizing each com-

ponent (dividing by 255), the result was converted

to the luminance component Y , computed as Y =

0.299R+0.587G+0.114B,where R, G, and B are the

red, green, and blue components, respectively, of the

colour image. Several other works on the detection of

retinal vessels have used the green channel only; in

the present work, the Y channel was used in order to

reduce noise by averaging the three colour component

images and obtain an image with positive contrast for

the vessels. The artifacts present in the DRIVE im-

ages at the edges were removed by applying morpho-

logical erosion with a disc-shaped structuring element

of diameter 10 pixels.

As the contrast between the blood vessels (fore-

ground) and the retinal tissue (background) is gener-

ally poor in the retinal images, an effective technique

called contrast-limited adaptive histogram equaliza-

tion (CLAHE) is utilized for contrast enhancement

by limiting the maximum slope in the transformation

function. Instead of applying the histogram equal-

ization on the entire image, it is applied only on

small non-overlapping regions in the image. Then,

the neighboring tiles are combined using bilinear in-

terpolation to reduce induced boundaries. Figure 1(b)

shows the contrast enhancementproducedby CLAHE

approach.

2.2 Blood Vessel Segmentation

In many applications of image processing in ophthal-

mology, the most important step is to detect the blood

vessels in the retina (Hoover et al., 2000),(Foracchia

et al., 2004). In our case, we decided to choose

the Soares et al. method (Soares et al., 2006) due

to his high performance in blood vessel segmenta-

tion (≈ 96%). The method produces segmentations

by classifying each image pixel as vessel or non-

vessel, based on the pixel’s feature vector. Feature

vectors are composed of the pixel’s intensity and two-

dimensionalGabor wavelet transformresponses taken

at multiple scales. Gabor wavelets are sinusoidally

modulated Gaussian functions that have optimal lo-

calization in both the frequency and space domains,

thus allowing noise ﬁltering and vessel enhancement

(a) (b)

Figure 1: (a) Green channel of image, (b) Result of contrast

enhancement with CLAHE.

in a single step. The wavelet is capable of detecting

directional structures and of being tuned to speciﬁc

frequencies, which is specially important for ﬁltering

out the background noise present in retinal images.

The 2-D Gabor wavelet is deﬁned as

(x) = exp( jk

x)exp



−0.5|Ax|



(1)

where A = diag





√

ε, 1



is a 2×2 diagonal matrix

that deﬁnes the anisotropy of the ﬁlter, i.e., its elon-

gation in any desired direction. The Gabor wavelet is

actually a complex exponential modulated Gaussian,

where k

is a vector that deﬁnes the frequency of the

complex exponential.

The Gabor wavelet transform is computed for

spanning from 0 up to 170 degrees at steps of 10 de-

grees. The maximum moduli of the wavelet trans-

form over all angles for various scales are then taken

as pixel features (Fig. 2(a)). In the tests performed,

the elongation parameter was set to ε = 4 and k

= 3.

The contrast enhancement image is inverted before

the application of the wavelet transform to it, so that

the vessels appear brighter than the background.

(a) (b)

Figure 2: (a) Maximum modulus response of Gabor wavelet

transform over 18 Gabor ﬁlters with scale value of a = 4. (b)

Segmentation of blood vessels.

The blood vessel segmentation is obtained using

a Bayesian classiﬁer with class conditional probabil-

ity density functions, including a Gaussian mixture

model, where each pixel is classiﬁed as a vessel or

non-vessel pixel, as shown in Fig. 2(b). To reduce

OPTIC DISC DETECTION IN RETINAL IMAGES BY PATTERN DISTANCE MINIMIZATION

357

the computational burden, pattern matching is applied

only to candidate pixels picked from the fundus im-

age. The binary vessel/nonvessel image is thinned by

applying a morphologic algorithm that reduces all ob-

jects in the binary image to the pixel dimension keep-

ing in the new image only the central pixels of the

vessels. The number of candidates is even more re-

duced by considering only the ones which are within

an image area with 2.5% of the pixels with the highest

intensity of the contracted image as showed in Fig. 3.

Figure 3: Thinned candidate vessels.

2.3 Optic Disc Detection

A distinguishing feature of the optic disc is that it

is the region of convergence for the blood vessel

network. The shape, colour, and size of the OD

showed large variance especially in the presence of

retinopathies, and therefore, detection methods based

on these properties were shown to be weak, and im-

practical.

As stated by Hoover et al. (Hoover et al., 2000),

”A matched ﬁlter describes the expected appearance

of a desired signal, for purposes of comparative mod-

eling”. Thus, in order to detect the OD, a simple ves-

sels’ direction matched ﬁlter is proposed to roughly

match the direction of the vessels at the OD vicin-

ity. Based on the retinal vasculature orientations we

deﬁne a 11×11 pattern formed by two opposite para-

bles. The distance between the pattern and the thinned

vessel map is obtained using the Earth Mover’s dis-

tance (EMD) approach.

Rubner et al. (Rubner et al., 2000) introduced the

EMD to measure perceptual similarity between im-

ages for the purpose of image retrieval. EMD evalu-

ates dissimilarity between two distributions or signa-

tures in some feature space where a distance measure

is given. Intuitively, given two distributions, one can

be seen as a mass of earth properly spread in space,

the other as a collection of holes. Then, the EMD

measures the least amount of work needed to ﬁll the

holes with earth. Here, a unit of work corresponds to

transporting a unit of earth by a unit of ground dis-

tance. The EMD between two distributions is given

Figure 4: Matching pattern.

by the minimal sum of costs incurred to move all the

individual points between the signatures.

Let P = {(p

, w

), ..., (p

, w

)} be the ﬁrst sig-

nature with m pixels, where p

is the pixel represen-

tative and w

is the weight of the pixel; the sec-

ond signature with n pixels is represented by Q =



, w

), ..., (q

, w

)



; and D = [d

] the distance

matrix where d

is the distance between two points’

image coordinates p

and q

. The ﬂow f

is the

amount of weight moved from p

to q

. The EMD

is deﬁned as the work normalized by the total ﬂow

, that minimizes the overall cost:

EMD(P, Q) =

∑

(2)

As pointed by Rubner et al. (Rubner et al., 2000),

if two weighted point sets have unequal total weights,

EMD is not a true metric. It is desirable for ro-

bust matching to allow point sets with varying total

weights and cardinalities. In order to embed two sets

of contour features with different total weights, we

divide the blood vessel image in several overlapping

tiles and simulate equal weights by adding the appro-

priate number of points, to the lower weight set, with

a penalty of maximal distance. As a measure of dis-

tance for the EMD, we use

= 1−e

−

−q

(3)

where



−q



is the Euclidean distance between p

and q

and α is used in order to accept some defor-

mation of the matching pattern. The exponential map

limits the effect of large distances, which otherwise

dominate the result.

3 RESULTS AND DISCUSSION

The proposed method was tested with fundus images

of the retina from the DRIVE database which contains

40 images (20 for training and 20 for testing).

BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing

358

(a) (b) (c) (d)

(e) (f) (g) (h)

Figure 5: OD detection examples (black mark represents the estimated OD centre). The ﬁrst row shows the best four results.

The second row shows the worst four results. Images (e) and (b) of second row are the only two failed detections.

Figure 5 shows some illustrative OD detection re-

sults. First row shows the best four results and the

second row shows the worst detections.

The proposed method achieved a success rate of

95% (i.e., the OD was detected correctly in 38 out of

the 40 images contained in the DRIVE database). The

estimated centre is considered correct if it was posi-

tioned within 60 pixels of the manually identiﬁed cen-

tre, as proposed in (Foracchia et al., 2004),(Hoover

et al., 2000),(Youssif et al., 2008). The average dis-

tance (for the 38 successful images) between the es-

timated OD centre and the manually identiﬁed center

was 24.3 pixels. The only two cases in which the OD

was not correctly detected (Images (e) and (f) of Fig.

5) were due to uneven crescent-shaped illumination

joint with crossed vessels that biased the OD candi-

dates and affected the vessel candidate selection.

4 CONCLUSIONS

In this paper, a simple and computationally efﬁ-

cient algorithm for automatic OD detection has been

presented. The proposed algorithm takes advan-

tage of the powerful preprocessing techniques such

as the contrast enhancement, Gabor wavelet trans-

form, mathematical morphology and Earth Mover’s

distance as the matching process. The performance

of the proposed algorithm has been evaluated using

DRIVE database images. Good results have been

achieved, failed the OD detection only in two of the

forty images.

REFERENCES

Foracchia, M., Grisan, E., and Ruggeri, A. (2004). Detec-

tion of optic disc in retinal images by means of a geo-

metrical model of vessel structure. IEEE Transactions

on Medical Imaging, 23(10):1189–1195.

Hoover, A., Kouznetsova, V., and Goldbaum, M. (2000).

Locating blood vessels in retinal images by piecewise

threshold probing of a matched ﬁlter response. IEEE

Transactions on Medical Imaging, 19(3):203–210.

Rubner, Y., Tomasi, C., and Guibas, L. (2000). The earth

mover’s distance as a metric for image retrieval. Inter-

national Journal of Computer Vision, 40(2):99–121.

Soares, J., Leandro, J., Jr., R. C., Jelinek, H., and Cree,

M. (2006). Retinal vessel segmentation using the 2-

d gabor wavelet and supervised classiﬁcation. IEEE

Transactions on Medical Imaging, 25(9):1214–1222.

Staal, J., Abramoff, M., Niemeijer, M., Viergever, M., and

van Ginneken, B. (2004). Ridge-based vessel segmen-

tation in color images of the retina. IEEE Transactions

on Medical Imaging, 23(4):501–509.

Youssif, A., Ghalwash, A., and Ghoneim, A. (2008). Optic

disc detection from normalized digital fundus images

by means of a vessels’ direction matched ﬁlter. IEEE

Transactions on Medical Imaging, 27(1):11–18.

OPTIC DISC DETECTION IN RETINAL IMAGES BY PATTERN DISTANCE MINIMIZATION

359