- •1. Introduction
- •2. Proposed retinal identification algorithm
- •2.1. Vessel Extraction
- •2.1.1. Preprocessing
- •2.1.2. Vessel Segmentation
- •2.2. Regions of Interest Definition and Extraction
- •2.2.1. Feature Extraction
- •2.2.1.1. Region-based Shape Features
- •2.2.1.2. Boundary-based Shape Features
- •2.2.1.3. Angle of sr Corners
- •2.2.1.4. Centroid Distance
- •2.2.1.5. Differential Tangent Angle
- •2.2.1.6. Weighted Corner Angle Feature
- •2.3. Proposed Hierarchical Matching Structure
- •2.3.1. Candidate Selection
- •2.3.2. Corner Matching
- •2.3.3. Sr Matching
- •2.4. Decision Making Scenario
- •3. Experimental Results
- •3.1. Database Description
- •3.2.1 Utilized Feature Order in Candidate Selection Step
- •3.2.2 Number of the Most Similar Candidate sRs
- •3.2.3. Parameter Setting of sr Matching
- •3.2.4. Parameter Setting of Decision Making Scenario
- •3.3. Identification Result Analysis
- •3.4. Discussion and Comparison with Other Works
- •4. Conclusions
2.3.2. Corner Matching
The goal of corner matching module is matching a point on the boundary of the query SR with its corresponding point on the boundary of the selected candidate SRs by the candidate selection module. For this purpose, we apply angle of SR corner and centroid distance features as it is described in the following.
Assume
be
the query SR and
be
one of
selected
candidates in the previous module which exhibit the least distance
from the query. Therefore, based on the SR corner angle feature,
and
would be sets of the extracted corner angles of query SR and enrolled
candidate SR, respectively. To match a corner point on
’s
boundary with its correspondence on
,
we propose an algorithm based on the least value of centroid distance
and corner angles as follows:
for
each candidate
for
for
if
end if
end for
if
Eliminate corresponding candidate
else
Specify the
coordinates on the boundary belonging to matched corners in the query
SR and its correspondence in the enrolled candidate
,
end if
end
end
where
are the extracted centroid distance of query SR and enrolled
candidate SR while their beginning points are
and
,
respectively and function
calculates Euclidean distance between two centroid distance vectors.
is
a constant that is set to 30 degrees in our application. With
applying this algorithm, a pair of points on two boundaries of the
query SR and candidate SR are determined as matched corners. As
described in our proposed corner point algorithm, the number of
candidates for further module can be reduced if the matching between
the corners of the query SR and candidate SR does not happen.
2.3.3. Sr Matching
The main goal of SR matching module is matching a SR among remaining candidates with query SR or rejecting the query SR. In Fig. 7, among remaining candidates (utmost SRs), we choose the most similar SR to query based on the similarity of differential tangent angle feature where their beginning points are the matched corner points obtained from corner matching module as,
(12)
where
and
are
differential tangent angle feature vectors of the query SR and
enrolled candidate, respectively and
is
the maximum number of remaining candidate SRs.
Now, the most similar enrolled SR in database with the query SR and also their matched corner points have been determined. In this step, we put a threshold to accept or reject the last candidate using weighted corner angle feature while the matched corner point is used to extract the feature from the query as follows,
(13)
where
,
are
the extracted weighted corner angle feature of query SR and enrolled
SR and
is a threshold to reject or accept the last remaining candidate.
