ivote3.m
4.79 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
function [px, py] = ivote3(image, rmax, phi, iter, t0, sigma, t1)
% This function performs iterative voting on an image in order to identify the centers of blobs
% image = a matrix of values representing a 2D image
% rmax = estimated radius of a blob
% phi = angular range of the voting area (in radians)
% iter = number of iterations for voting
% sigma = standard deviation of the weighting filter used for voting
% t0 = initial gradient magnitude threshold (limits the # of voting pixels)
% t1 = final threshold for valid points
dphi = phi / (iter);
%compute the gradient and gradient magnitude
[Igrad_x, Igrad_y] = gradient(image);
Igrad_mag = sqrt(Igrad_x.^2 + Igrad_y.^2);
%represent the gradient direction in polar coordinates
Igrad_theta = atan2(Igrad_x,Igrad_y) + pi;
%compute pixels that will be processed
% calculate a mask representing the voting pixels
S = Igrad_mag > t0;
Sx = size(S, 1);
Sy = size(S, 2);
S(1:rmax, :) = 0;
S(Sx - rmax:Sx, :) = 0;
S(:, 1:rmax) = 0;
S(:, Sy - rmax:Sy) = 0;
%calculate the coordinates of all of the voting pixels
[sx, sy] = find(S); %this provides a coordinate for each voter
%eliminate pixels close to the image boundary
nV = nnz(S); %calculate the number of valid pixels (voters)
% create a meshgrid describing coordinates relative to the voter
% position
range = -rmax:rmax;
[Vx, Vy] = meshgrid(range, range);
%create a mask for the voting area
M_dist = Vx.^2 + Vy.^2 < rmax^2;
M_theta = atan2(Vx, Vy);
pxPerRow = size(image,1);
g_v_prime = zeros(nV, ceil(rmax^2*phi/3));
validPoints = zeros(nV,1);
for i = 1: iter
t_iter = tic;
%set the vote image to zeros
Ivote = zeros(size(image));
for v = 1:nV
%determine the position of the current voter
vx = sx(v);
vy = sy(v);
%determine the orientation (theta) of the current voter
vtheta = Igrad_theta(vx, vy);
% find the angular distance between M_theta and V_theta at each
ang_diff = abs(M_theta - vtheta);
M_diff = min(2*pi - ang_diff,ang_diff)<phi;
%compute the final mask
M = (M_dist .* M_diff);
%get the coordinates of each pixel in the final mask
[vx_prime, vy_prime] = find(M);
%transform the local coordinates of the pixels in the voting region
%to the global coordinates of the image
npts = numel(vx_prime);
validPoints(v) = npts;
g_v_prime(v,1:npts) = vx + (vx_prime - (rmax + 1)) + (vy + (vy_prime - (rmax + 1))-1).*pxPerRow;
% compute the Gaussian kernel for the current voter
%c_x = (max(vx_prime) + min(vx_prime))/2
%c_y = (max(vy_prime) + min(vy_prime))/2
c_x = rmax/2 * cos(vtheta)+rmax;
c_y = rmax/2 * sin(vtheta)+rmax;
Gauss_size_x = vx_prime -c_x;
Gauss_size_y = vy_prime -c_y;
gauss_p1 = 1/(2*pi*sigma^2);
gauss_p2 = -1/(2*sigma^2);
Gauss_p3 = (Gauss_size_x.^2 + Gauss_size_y.^2);
gaussian = gauss_p1*exp(gauss_p2*Gauss_p3);
% *******************************************
%determine the gradient magnitude at the current voter location
vmag = Igrad_mag(vx, vy);
%add the voter's current magnitude weighted by gaussian function to the vote image
for n=1: npts
% Ivote( g_v_prime(v,n)) = Ivote( g_v_prime(v,n)) + gaussian(n)*vmag;
Ivote( g_v_prime(v,n)) = Ivote( g_v_prime(v,n)) + vmag;
end
end
%update the voting direction for each voter
for v = 1:nV
%determine the position of the current voter
vx = sx(v);
vy = sy(v);
%get the local value of the voting image
local_Ivote = Ivote(g_v_prime(v,1:validPoints(v)));
%find the index of the maximum value
[~, local_max_idx] = max(local_Ivote);
%convert this into a global subscript
gx = rem(g_v_prime(v,local_max_idx)-1, pxPerRow) + 1;
gy = (g_v_prime(v,local_max_idx) - gx)/pxPerRow + 1;
%compute the vector from the voter position to this position
new_vx = gx - vx;
new_vy = gy - vy;
%convert this to a new theta value
new_vtheta = atan2(new_vy, new_vx);
%update the gradient image with the new voter direction
Igrad_theta(vx, vy) = new_vtheta;
end
%reduce phi for the next iteration
phi = phi - dphi;
toc(t_iter)
end
% add threshold and find the local maxima
cell_center = Ivote;
cell_center (cell_center <t1) = 0;
cell_center = imregionalmax(cell_center);
imshow(cell_center);
% Create an Overlay
output = mat2gray(image);
output(:, :, 2) = mat2gray(cell_center);
output(:, :, 3) = mat2gray(image);
figure; imagesc(output);
imwrite(output, 'output image.bmp');
[px,py] = find(cell_center);