Figure 1: Eye Fundus Image
Detection and Localization of Fovea in Human
Retinal Fundus Images
Khawaja Muhammad Asim
1
, A. Basit
2
, Abdul Jalil
1
1
DCIS. Pakistan Institute of Engineering and Applied Sciences (PIEAS),
Nilore, Islamabad 45650, Pakistan
2
TPPD. Pakistan Institute of Nuclear Science and Technology (PINSTECH),
Nilore, Islamabad 45650, Pakistan
asimkhawaja786@gmail.com, abdulbasit1975@gmail.com, jalil@pieas.edu.pk
Abstract— In retinal images, Optic Disc (OD) and fovea are
two main structures. To develop an automated system for
detection of eye diseases like glaucoma and macular edema, it
is necessary to have knowledge of OD and fovea. This paper
describes an automated detection of fovea. Detection of fovea is
dependent upon approximate location of OD. After obtaining
the location of OD, erosion and dilation operations are
performed on gray scale fundus images. The region of interest
is defined with respect to the location of OD. Since in the case
of normal retina image fovea is the darkest spot inside macular
region and morphological operations make it more visible than
its macula background. So the set of all the darkest spots are
detected and their median point is selected as center of fovea.
This detected location is then mapped in blood vessel map of
retina. If it lies in the vicinity of any blood vessel then image
shows signs of macular edema and take the second minimum
intensity value for fovea location.
Keywords-component; Fovea detection, Fovea loclization,
Retinal images, Optic Disc
I. INTRODUCTION
While examining the normal retinal image, it is found
that fovea is a dark portion present on temporal side of
Optic Disc (OD) inside macula region as shown in Figure 1.
Center of fovea is away from OD about 2-3 times the
diameter of OD [1]. Fovea contains cones (light receptors)
in its central region, which are responsible for sharp vision.
Endeavors are on their way to design automated systems for
detection of eye diseases like glaucoma, diabetic retinopathy
and macular edema. This research is a small part of such big
project. Fovea is harder to detect as compare to OD because
it is the smallest part present inside macula region and it has
no clear boundary unlike OD. The main properties of fovea
that have been used by researchers for its detection are its
relative position with respect OD, absence of major blood
vessels inside it and its dark color. Several methods [1-5]
are available in literature for detection of fovea.
This paper presents a precise and fast method for
detection of fovea. However it needs the approximate
location and diameter of OD as input parameters. We
incorporate this by using techniques proposed by Mahfouz
et al. [6] and Basit [7]. Proposed method for fovea detection
is quite tolerant towards certain abnormalities of retina i.e.
lesions and hard exudates. Sinthanayothin et al. [4] in 1999
proposed a method for detection of fovea using a template
to find the fovea locus in retinal images. Singh et al. [2]
detected fovea using appearance based method. Sekhar et al.
[5] used Hough transform and morphological operators to
detect fovea in 2008. Most of the methods (available in
literature) tend to use the location of OD and blood vessels.
Therefore, they do not work properly if the accurate
information of OD location and blood vessels is not
available.
Method proposed in this paper also works accurately if
exact location of OD is not known. The only prerequisite is
the approximate location of OD and its approximate
diameter. If the image possess signs of macular edema at its
acute stage, then blood vessels map is also required for
accurate detection. During fovea detection entire blood
vessels from retinal image are removed using morphological
operators. Applying contrast enhancement makes the
intensity differences more prominent. Using the
approximate coordinates of OD point as a reference, a
region of interest (ROI) is defined and then fovea is detected
as the darkest portion in the region of interest. If accurate
values of OD coordinates or diameter are not known, a wide
Optic Disc
Fovea
978-1-4673-4451-7/12/$31.00 ©2012 IEEE