![](bg1.png)
Focus Evaluation Approach for Retinal Images
Diana Veiga
1,2
, Carla Pereira
1
, Manuel Ferreira
1,2
Luís Gonçalves
3
and João Monteiro
2
1
ENERMETER, Parque Industrial Celeirós 2ª Fase, Lugar de Gaião, Lotes 5/6, 4705-025 Braga, Portugal
2
Centro Algoritmi, University of Minho, Azurém, 4800-058 Guimarães, Portugal
3
Oftalmocenter, Rua Francisco Ribeiro de Castro, nº 205, Azurém, 4800-045 Guimarães, Portugal
Keywords: Digital Fundus Photography, Focus Measures, Image Processing.
Abstract: Digital fundus photographs are often used to provide clinical diagnostic information about several
pathologies such as diabetes, glaucoma, macular degeneration and vascular and neurologic disorders. To
allow a precise analysis, digital fundus image quality should be assessed to evaluate if minimum
requirements are present. Focus is one of the causes of low image quality. This paper describes a method
that automatically classifies fundus images as focused or defocused. Various focus measures described in
literature were tested and included in a feature vector for the classification step. A neural network classifier
was used. HEI-MED and MESSIDOR image sets were utilized in the training and testing phase,
respectively. All images were correctly classified by the proposed algorithm.
1 INTRODUCTION
Eye fundus imaging allows the observation of the
retina and the analysis of its constituents. With this
medical imaging examination several pathologies
can be diagnosed, mainly those related with blood
vessels modifications. In recent years there have
been numerous research attempts for the
development of systems to automatically analyze
fundus images. The success of these systems is
frequently affected by image quality which
sometimes is poor due to bad acquisition conditions
or the presence of occlusions, cataracts and opacities
in patients’ eyes. For a proper automated analysis,
fundus images must present a minimum quality that
not always is possible to guarantee by clinicians in
the capturing moment. Focus is one of the
parameters responsible for a reduced quality image,
which we propose to verify in digital fundus
photography.
The task of eye fundus image acquisition
demands a specific training as numerous conditions
must be fulfilled. Moreover, despite some
commercial fundus cameras comprise tools to assist
the photographer in the operation, focusing on the
fundus can be difficult and subjective.
Focus measures appear as methods to estimate
the sharpness of an image. Various algorithms have
been proposed for auto-focusing, estimating depth,
or just to determine the degree of blurring (Marrugo,
2012); (Yap, 2004); (Yang, 2003); (Pertuz, 2013)
(Moscaritolo, 2009). Pertuz et al., (2013) divides the
most popular measures in different groups:
Gradient-based operators, Wavelet-based operators,
Statistic-based operators, DCT-based operators and
Miscellaneous operators. However, very few of
these methods have been tested in fundus images
(Marrugo, 2012).
In general, a single focus operator is applied to
an image. Nonetheless, since fundus images content
extremely varies, a single focus operator cannot
always achieve a correctly focus estimation. To
address this issue, in this work, a group of focus
measures were selected and combined to be used in
a neural network classifier. A new approach to
automatically classify retinal images as
focused/defocused is described. Several experiments
were carried out using real focused fundus images
and synthetically defocused ones. Numerous focus
operators were tested and applied on the referred
digital images and their response to blur was
evaluated. In addition, this study reports the
application of an artificial neural network classifier
to obtain the final classification of retinal images.
Three focus measures were considered as input
features to the classifier: a wavelet-based measure, a
456
Veiga D., Pereira C., Ferreira M., Gonçalves L. and Monteiro J..
Focus Evaluation Approach for Retinal Images.
DOI: 10.5220/0004671104560461
In Proceedings of the 9th International Conference on Computer Vision Theory and Applications (VISAPP-2014), pages 456-461
ISBN: 978-989-758-003-1
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)