- •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
- •2.2.1.3. Angle of sr Corners
- •2.2.1. Feature Extraction
- •2.2.1.1. Region-based Shape Features
- •2.2.1.2. Boundary-based Shape Features
- •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
2.2.1.4. Centroid Distance
In the shape-based object recognition application, the centroid is
defined as the gravity center of the object [34]. This feature is
defined based on the distance of boundary points and the centroid
of
the selected SR as,
(5)
where
is the coordinate of each boundary pixel and
is calculated as,
,
(6)
where
represents
the total number of points that are belong to the selected SR
consisting of the inside and boundary points. Fig. 5 demonstrates the
extraction of this feature on a sample of the selected SR.
2.2.1.5. Differential Tangent Angle
Differential tangent angle feature is a strong feature to describe the curvature of a SR’s boundary. This feature provides the finer details of slope changes on the boundary of a SR which can be beneficial in similarity matching. To extract this feature from the softened boundary of selected SR, we define the sequence of slope of consecutive points as,
,
,
(7)
where
,
(8)
and represents the coordinates of points on the softened boundary (as it is described in Eq. (4)) . To make this feature more resistance against the rotation, we apply a 1-D Fourier transform on the extracted feature vector and use the absolute value of coefficients of Fourier transform as the differential tangent angle feature. This procedure is shown in Fig. 5.
(a)
(b)
(c)
(d)
Fig. 5: Extraction process of the centroid distance and differential tangent angle features. (a) Softened boundary of the selected SR in Fig. (3-a) and description of the centroid distance and differential tangent angle features, (b) Centroid distance vector of (a), (c) Differential tangent angle vector of (a), (d) 1-D Fourier Transform of (c).
2.2.1.6. Weighted Corner Angle Feature
To extract the rotation-invariant shape features from the SR, we propose a novel feature that can incorporate simultaneously variations of the curvature and shape of the SR object. For this purpose, for a corner point of the SR, the angle between two points of boundary in the left and right sides of that corner point is weighted with Euclidean distance of those two points. Therefore, valuable properties of the curvature of boundary points and shape of the SR based on a special corner point are integrated in a feature that is called weighted corner angle feature.
Suppose
is
a closed set of boundary points of a selected SR and
is
an extracted corner point on the selected SR’s boundary. Then, we
compute the angle between each pair of its neighbor points (right
handed and left handed) as,
(9)
where
is
the number of boundary points belong to selected SR and
and
are
the
left-handed
and right-handed neighborhood pixels of the corner point
,
respectively. Then, Euclidean distance between each pair of these
neighborhood points is computed as in the following,
(10)
To integrate two terms of the angle and distance in the weighted corner angle feature, we normalize the angle term in range of [0, 1] by sinusoidal function and then multiply this normalized term to distance term as,
(11)
Since each SR of a query image has more than one corner point, the scenario of corner point selection will be described in section (2.3.1). Fig. 6 illustrates the process of weighted corner angle feature for one corner point of the selected SR in Fig. (3-a).
(a)
(b)
(c)
(d)
Fig. 6: Extraction process of
weighted corner angle feature for one corner point of the selected SR
in Fig. (3-a). (a) Softened boundary of the selected SR in Fig. (3-a)
and description of the weighted corner angle feature, the green point
shows one of selected corner points as a sample, blue vectors
represent each pair of left and right-handed neighbors of selected
corner point used in Eq. 9 to compute angles between each two
vectors, red lines show the distances between each pair of left and
right-handed neighbors of selected corner point described in Eq. 10,
(b) Angle vector of
,
(c) Distance vector of
,
(d) Weighted corner angle feature
.
