基于matlab对比度和结构提取的多模态解剖图像融合实现

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
clear; close all; clc; warning off
%% A Novel Multi-Modality Anatomical Image FusionMethod Based on Contrast and Structure Extraction
% F = fuseImage(I,scale)

%Inputs:
%I - a mulyi-modal anatomical image sequence

%scale - scale factor of dense SIFT, the default value is 16

%% load images from the folder that contain multi-modal image to be fused
%I=load_images('./DatasetCT-MRIPair 1');
I=load_images('./DatasetMR-T1-MR-T2Pair 1');
%I=load_images('./DatasetMR-Gad-MR-T1Pair 1');
% Show source input images 
figure;
no_of_images = size(I,4);
for i = 1:no_of_images
    subplot(2,1,i); imshow(I(:,:,:,i));
end
suptitle('Source Images');


%%
F=fuseImage(I,16);
%% Output: F - the fused image

F=rgb2gray(F);
figure;
imshow(F);
function [ F ] = fuseImage(I,scale)


addpath('Pyramid_Decomposition');
addpath('Guided_Filter');
addpath('Dense_SIFT');

tic
%%
[H, W, C, N]=size(I);
imgs=im2double(I);
IA=zeros(H,W,C,N);
for i=1:N
IA(:,:,:,i)=enhnc(imgs(:,:,:,i));

end
%%
imgs_gray=zeros(H,W,N);
for i=1:N
    imgs_gray(:,:,i)=rgb2gray(IA(:,:,:,i));
end
%
% %dense sift calculation
dsifts=zeros(H,W,32,N, 'single');
for i=1:N
    img=imgs_gray(:,:,i);
    ext_img=img_extend(img,scale/2-1);
    [dsifts(:,:,:,i)] = DenseSIFT(ext_img, scale, 1);
    
end
%%
%local contrast
contrast_map=zeros(H,W,N);
for i=1:N
    contrast_map(:,:,i)=sum(dsifts(:,:,:,i),3);

end


%winner-take-all weighted average strategy for local contrast

[x, labels]=max(contrast_map,[],3);
clear x;
for i=1:N
    mono=zeros(H,W);
    mono(labels==i)=1;
    contrast_map(:,:,i)=mono;

end



%% Structure 
h = [1 -1];
structure_map=zeros(H,W,N);

for i=1:N
structure_map(:,:,i) = abs(conv2(imgs_gray(:,:,i),h,'same')) + abs(conv2(imgs_gray(:,:,i),h','same')); %EQ 13

   
end


%winner-take-all weighted average strategy for structure

[a, label]=max(structure_map,[],3);
clear x;
for i=1:N
    monoo=zeros(H,W);
    monoo(label==i)=1;
    structure_map(:,:,i)=monoo;
     
end

%%
weight_map=structure_map.*contrast_map;




%weight map refinement using Guided Filter
for i=1:N
    
    weight_map(:,:,i) = fastGF(weight_map(:,:,i),12,0.25,2.5);
 
end



% normalizing weight maps
%
weight_map = weight_map + 10^-25; %avoids division by zero
weight_map = weight_map./repmat(sum(weight_map,3),[1 1 N]);

%% Pyramid Decomposition

% create empty pyramid
pyr = gaussian_pyramid(zeros(H,W,3));
nlev = length(pyr);

% multiresolution blending
for i = 1:N
    % construct pyramid from each input image
   
    % blend
    for b = 1:nlev
        w = repmat(pyrW{b},[1 1 3]);
        
        pyr{b} = pyr{b} + w .*pyrI{b};
    end
    
end

% reconstruct
F = reconstruct_laplacian_pyramid(pyr);

toc

end