Contents lists available at ScienceDirect
Pattern Recognition
journal homepage: www.elsevier.com/locate/pr
Optic disc segmentation based on variational model with multiple energies
Baisheng Dai
a
, Xiangqian Wu
a,
⁎
, Wei Bu
b
a
School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China
b
Department of New Media Technologies and Arts, Harbin Institute of Technology, Harbin 150001, China
ARTICLE INFO
Keywords:
Retinal disease
Optic disc segmentation
Optic disc localization
Sparse coding
Variational model
ABSTRACT
Accurate and reliable optic disc (OD) segmentation is important for retinal image analysis and retinal disease
screening. This paper presents a novel method to automatically segment OD in fundus images based on
variational model with multiple energies. Firstly, a sparse coding based technique is designed to localize the OD
center, based on which an initial boundary curve is then estimated by a circular Hough transform. Next, OD
segmentation is regarded as an energy minimization problem, and a variational model integrating three energy
terms is proposed to evolve the curve to the OD boundary. In the proposed model, the first term, named phase-
based boundary energy, is designed to attract the evolution curve to the OD boundary, even the one with low
contrast; the second term, named PCA-based shape energy, constraints the evolution curve to a common OD
shape, which can suppress the negative effect of bright interferences, e.g., the bright lesions and myelinated
nerve fibers, in OD segmentation; the last one is the region energy, which drives the evolution curve to the
boundary of the homogeneous regions and hence improve the robustness of the model to the noises boundary
and the initial position of evolution curve. The proposed OD segmentation method is evaluated on three public
available databases, i.e., the MESSIDOR, ONHSD and DRIONS databases, and the experimental results
demonstrate that the proposed method outperforms the state-of-the art techniques.
1. Introduction
The optic disc (OD) is one of the main anatomical structures in
retinal images. The OD generally appears as an approximately circular
and bright yellowish object in normal retinal image [1], and it is also
the entry point of the major blood vessels that supply the retina [2],as
shown in Fig. 1.
Accurate OD segmentation plays an important role in retinal image
analysis and automated screening for fundus diseases. The OD location
can be used as a reference for fovea detection, vessel tracking and
measurement and other tasks [3,4]. The segmented OD provides the
diagnostic information for automated eye diseases screening, such as
glaucoma, papilledema, neovascularization of the disc (NVD), and
hypertensive retinopathy [5–7].
Although an OD has well-defined features in a normal retinal
image, OD segmentation is still very challenging in abnormal retinal
images. In past decades, many techniques of automatic OD segmenta-
tion have been investigated. Most existing OD segmentation techniques
could be roughly grouped into three categories: template-based tech-
niques, morphology-based techniques and deformable model-based
techniques.
Template-based techniques are mainly based on the shape prior of
the OD, i.e. the circular or elliptical shape. Pinz et al. [8], Aquino et al.
[9], and Lalonde et al. [10] used a Hough transform to determine
boundaries of an OD. Wong et al. [11], Roychowdhury et al. [12] and
Morales et al. [6] applied an ellipse or a circle to fit the OD boundary.
However, OD boundaries are not strictly circle or ellipse, and these
methods may miss some part of real OD boundaries.
Morphology-based techniques, which primarily make use of bright-
ness and shape property of OD [13–15], detect the boundaries of the
OD by utilizing morphological operators to find an isolated circular
bright region. However, these techniques are likely to be disturbed by
bright lesions (e.g., exudates and cotton-wool spots) in fundus images
which may share the similar appearance with OD, and sometimes,
mistakenly find the boundary of the optic cup, a cup-like area at OD
center, which always have higher brightness than OD.
To acquire more accurate boundary of the OD, deformable model-
based techniques have been largely adopted for OD segmentation.
Lowell et al. [2] segmented OD boundary with a global elliptical
parametric model combined with a local deformation model. Li and
Chutatape [16] proposed a modified active shape model to segment
ODs, where PCA is first performed to build the shape model of ODs
using training images, and the OD boundary is then detected by
searching an instance of such shapes in test images. Xu et al. [5]
http://dx.doi.org/10.1016/j.patcog.2016.11.017
Received 1 August 2016; Received in revised form 10 October 2016; Accepted 17 November 2016
⁎
Corresponding author.
E-mail address: xqwu@hit.edu.cn (X. Wu).
Pattern Recognition 64 (2017) 226–235
Available online 20 November 2016
0031-3203/ © 2016 Elsevier Ltd. All rights reserved.
MARK