Semantic-Filtering/SemanticFilter.m at master · Dullson/Semantic-Filtering · 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
function [F, E] = SemanticFilter(img, sigma_s, sigma_r, edge_model, num_iterations, joint_image)

    %% Setting up the arguments
    I = double(img);
    if ~exist('num_iterations', 'var')
        num_iterations = 3;
    end

    if exist('joint_image', 'var')
        J = double(joint_image);

        if (size(I,1) ~= size(J,1)) || (size(I,2) ~= size(J,2))
            error('Input and joint images must have equal width and height.');
        end
    else
        J = I;
    end

    [h, w, num_joint_channels] = size(J);


    %% Perform the filtering.
    N = num_iterations;
    F = I;

    sigma_H = sigma_s;

    for i = 0:num_iterations - 1

    % Apply Median filter
    % mfs   : median frame size
    mfs = 5;
    if mfs > 1

        M = medfiltimg(F, [mfs mfs]);
        mfs = mfs-2;
    else
        M=F;
    end

    % Estimate horizontal and vertical partial derivatives using finite
    % differences.
    dIcdx = diff(M, 1, 2);
    dIcdy = diff(M, 1, 1);

    dIdx = zeros(h,w);
    dIdy = zeros(h,w);

    % Compute the l1-norm distance of neighbor pixels.
    for c = 1:num_joint_channels
        dIdx(:,2:end) = dIdx(:,2:end) + abs( dIcdx(:,:,c) );
        dIdy(2:end,:) = dIdy(2:end,:) + abs( dIcdy(:,:,c) );
    end


        % detect edge probabilities and get map
        E = edgesDetect(F,edge_model);

        % Compute the derivatives of the horizontal and vertical domain transforms.
        dHdx = (1 + sigma_s/sigma_r * (E .* dIdx));
        dVdy = (1 + sigma_s/sigma_r * (E .* dIdy))';

        % Compute the sigma value for this iteration
        sigma_H_i = sigma_H * sqrt(3) * 2^(N - (i + 1)) / sqrt(4^N - 1);

        F = TransformedDomainRecursiveFilter_Horizontal(F, dHdx, sigma_H_i);
        F = image_transpose(F);

        F = TransformedDomainRecursiveFilter_Horizontal(F, dVdy, sigma_H_i);
        F = image_transpose(F);

    end

    F = cast(F, class(img));

end

%% Recursive filter.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function F = TransformedDomainRecursiveFilter_Horizontal(I, D, sigma)

    % Feedback coefficient
    a = exp(-sqrt(2) / sigma);

    F = I;
    V = a.^D;

    [h, w, num_channels] = size(I);

    % Left -> Right filter.
    for i = 2:w
        for c = 1:num_channels
            F(:,i,c) = F(:,i,c) + V(:,i) .* ( F(:,i - 1,c) - F(:,i,c) );
        end
    end

    % Right -> Left filter.
    for i = w-1:-1:1
        for c = 1:num_channels
            F(:,i,c) = F(:,i,c) + V(:,i+1) .* ( F(:,i + 1,c) - F(:,i,c) );
        end
    end

end

%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function T = image_transpose(I)

    [h, w, num_channels] = size(I);

    T = zeros([w h num_channels], class(I));

    for c = 1:num_channels
        T(:,:,c) = I(:,:,c)';
    end

end

function M = medfiltimg(I, kernel)
    [h, w, num_channels] = size(I);

    M = zeros([h w num_channels], class(I));

    for c = 1:num_channels
        M(:,:,c) = medfilt2(I(:,:,c), kernel);
    end

end