Page 141 - Computational Retinal Image Analysis

P. 141

CHAPTER

8
Image quality assessment

a
a
b
Sarah A. Barman , Roshan A. Welikala , Alicja R. Rudnicka ,
b
Christopher G. Owen
a School of Computer Science and Mathematics, Kingston University,
Kingston upon Thames, United Kingdom,
b Population Health Research Institute, St. George’s University of London, London,
United Kingdom

1 Introduction
The application of automated analysis techniques to large numbers of ophthalmic
images to retrieve useful clinical information is becoming more widespread due to
increasing use of routine fundus imaging, availability of datasets and the emergence
of increasingly sophisticated analysis techniques. Applications range from population
screening for ophthalmic disease detection to epidemiological studies which seek to
link retinal morphometric measurements to disease risk and processes. The performance
of automated image analysis systems on large sets of images is often dependent on the
quality of the image being assessed and this important aspect is reflected in a growing
scientific literature which focuses on image quality assessment (IQA) algorithms and
their use.
The scope of the main sections of this chapter will be based on IQA algorithms
applied to retinal fundus camera images. The applications covered are limited to
clinical examples, and biometric applications are not included. The first section will
describe how retinal fundus images can vary in quality. Example applications will
also be summarized to give an overview of the main challenges. An overview of
different algorithms that have been employed to assess image quality and the datasets
and metrics used for evaluation will be given in the second section. Summaries of
selected algorithms to highlight particular methods and applications follow. The final
concluding section includes comment on IQA techniques related to other imaging
modalities such as optical coherence tomography.

136 137 138 139 140 141 142 143 144 145 146