Fringe noise removal of retinal fundus images using trimming regions - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 53

Based on the retinal images and their characteristics, either the circular or elliptical trim-
ming region may be used and implemented.
2.1 Implementation
The implementation of the proposed methodology is similar to what has been previously sug-
gested by the authors [ 26 ] with minor modifications. The needed information for both the cir-
cular and elliptical trimming region is the estimated location for the center and the radius in
both horizontal and vertical directions.
Initial step would be creating a mask to define the region of interest [ 27 ] . Once the region
of interest is defined, the radius can then be estimated. To estimate the radius, it is important
to locate the first and last nonzero pixel (white pixel) across the columns of the image. The
middle value between these two estimated points can then be used as a preliminary location
for the center.
This is then followed by finding the first and last nonzero pixel across the rows using the
estimated center location. Recalculating the middle value of these newly found points can lead
to readjustment of the center location as well as the location of the first and last nonzero pixels
across the columns of the image.
In order to find the radius, the distance between the radius and the corner pixel values is
calculated. For the circular trimming region, both the vertical and horizontal calculated radius
would approximately be the same and so either can be used to define the trimming region.