Transcribing_Text_Museum_Labels/matchDaisyFeatures.m at master · nitinagarwal/Transcribing_Text_Museum_Labels · GitHub

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
function matchedpts = matchDaisyFeatures(featuresDaisy_im1,featuresDaisy_im2,dzy1,dzy2,im1,im2)

radius = 20; % was 20
threshold = 1.0;
matchpts_im1=[];
matchpts_im2=[];
score = [];

fvector1 = featuresDaisy_im1.vector;
fvector2 = featuresDaisy_im2.vector;
l1 = featuresDaisy_im1.location;
l2 = featuresDaisy_im2.location;

for i=1:size(fvector1,1)

    if(l1(i,1)<radius+1 || l1(i,2)<radius+1)
        continue;
    end

    [X,Y]=meshgrid(l1(i,1)-radius:l1(i,1)+radius,l1(i,2)-radius:l1(i,2)+radius); % computing via neighbourhood
    p=[X(:) Y(:)];
    features = extractDaisyFeatures(dzy2,p);
    result1 = bsxfun(@minus,features.vector,fvector1(i,:));  %eucledian distance between feature vector
    result1 = sqrt(sum(result1.^2,2));
    id1 = find(result1==min(result1));

    if(numel(id1)>1)
        disp('id1');
%         break;
        [id1,~] = findBetterNCC(l1(i,:),features.location(id1,:),im1,im2);
%         [id1,~] = findClosestPoint(l1(i,:),features.location(id1,:));       % closest pt if tie
    end

    result2 = bsxfun(@minus,fvector2,fvector1(i,:));  %eucledian distance between feature vector
    result2 = sqrt(sum(result2.^2,2));
    id2 = find(result2==min(result2));

    if(numel(id2)>1)
        disp('id2');
%         break;
        [id2,~] = findBetterNCC(l1(i,:),features.location(id1,:),im1,im2);
%         [id2,~] = findClosestPoint(l1(i,:),featuresDaisy_im2.location(id2,:)); % closest pt if tie
    end

    if(result1(id1) < result2(id2))

            matchpts_im1 = [matchpts_im1; l1(i,:)];
            matchpts_im2 = [matchpts_im2;features.location(id1,:)];
            score = [score;result1(id1)];

    elseif(result1(id1) > result2(id2))

        matchpts_im1 = [matchpts_im1; l1(i,:)];
        matchpts_im2 = [matchpts_im2;featuresDaisy_im2.location(id2,:)];
        score = [score;result2(id2)];

    elseif(result1(id1) == result2(id2))

           disp('score is equal');
%            break;
           [value,~] = findBetterNCC(l1(i,:),[features.location(id1,:); featuresDaisy_im2.location(id2,:)],im1,im2);
%          [value,~] = findClosestPoint(l1(i,:),[features.location(id1,:); featuresDaisy_im2.location(id2,:)]);

           if(value==1)
               matchpts_im1 = [matchpts_im1; l1(i,:)];
               matchpts_im2 = [matchpts_im2;features.location(id1,:)];
           else
               matchpts_im1 = [matchpts_im1; l1(i,:)];
               matchpts_im2 = [matchpts_im2;featuresDaisy_im2.location(id2,:)];
           end
           score = [score;result1(id1)];
    end
%     pause
%     figure,imshow(im1);hold on
%     plot(l1(i,1),l1(i,2),'b*','MarkerSize',3);
%     figure,imshow(im2);hold on
%     plot(matchpts_im2(end,1),matchpts_im2(end,2),'r*','MarkerSize',3);

end

if(size(matchpts_im1,1) <5)
    matchedpts.score = nan;
    matchedpts.image1=nan;
    matchedpts.image2=nan;
    return;
end


% outlier removal (ransac)
% assuming there is an affine transformation between two images using just
% these small correspondences.
% [~, inlier_im1,inlier_im2] = estimateGeometricTransform(matchpts_im1,matchpts_im2,'affine','MaxDistance',20);

% sprintf('size of inliers is %i',size(inlier_im1,1))
% [inlier_im1,inlier_im2] = removeCorrespondenceOutliers(inlier_im1,inlier_im2,im1,im2);
[inlier_im1,inlier_im2] = removeCorrespondenceOutliers(matchpts_im1,matchpts_im2,im1,im2);
sprintf('size of modified inliers is %i',size(inlier_im1,1))

if(isempty(inlier_im1))
    matchedpts.score = nan;
    matchedpts.image1=nan;
    matchedpts.image2=nan;
    return;
end

[~,~,ib]=intersect(inlier_im1,matchpts_im1,'rows');
newscore = score(ib);

% sprintf('size of inliers is %i',length(newscore));
% try distribute the pts evenly or randomly
% numpts = round(length(newscore)/10);
numpts=1;
matchedpts.score = newscore(1:numpts:end);
matchedpts.image1=inlier_im1(1:numpts:end,:);
matchedpts.image2=inlier_im2(1:numpts:end,:);


% attempt 1
%taking 10 pts with the lowest scores
% [score,id]=sort(newscore,'descend');

% matchedpts.score = score(1:10);
% matchedpts.image1=inlier_im1(id(1:10),:);
% matchedpts.image2=inlier_im2(id(1:10),:);

% attempt 2
% %based on distance
% dis = sqrt(sum((matchpts_im1 - matchpts_im2).^2,2));
%
% id = find(dis<100);
%
% a = matchpts_im1(id,:);
% b = matchpts_im2(id,:);


end