Fringe noise removal of retinal fundus images using trimming regions - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 52

The above procedure improved the precision of detection greatly and if the chosen radius
was accurately selected, the crucial information would have still been intact. Despite the great
success rate of this method, it can only be implemented on retinal images where the re-
gion of interest is approximately circular. Observing multiple retinal fundus images suggest
that diferent instrumentations and setings may also result in an oval-shaped retinal images.
Examples of which are shown in Figure 3 .
FIGURE 3 Circular- [ 18 ] and oval-shaped retinal fundus images [ 19 , 20 ] .
In these cases implementing the circular trimming region may result in an inaccurate detec-
tion of the feature of interest as well as great loss of crucial information, as illustrated in Figure
4 .
FIGURE 4 Implementation of circular trimming region to the oval-shaped retinal image.
To overcome this problem and preserve as much information as possible, authors propose
another trimming region. The suggested elliptical trimming region is characterized by the fol-
lowing equations:
(3)