Low-light images are not conducive to human observation and computer vision algorithms due to their low visibility. Although many image enhancement techniques have been proposed to solve this problem, existing methods inevitably introduce contrast under- and over-enhancement. In this paper, we propose an exposure fusion framework and an enhancement algorithm to provide an accurate contrast enhancement. Specifically, we first design the weight matrix for image fusion using illumination estimation techniques. Then we introduce our camera response model to synthesize multi-exposure images. Next, we find the best exposure ratio so that the synthetic image is well-exposed in the regions where the original image under-exposed. Finally, the input image and the synthetic image are fused according to the weight matrix to obtain the enhancement result. Experiments show that our method can obtain results with less contrast and lightness distortion compared to that of several state-of-the-art methods.

1
2
3
4
5
6
7
8
@inproceedings{ying2017new,
title={A New Image Contrast Enhancement Algorithm Using Exposure Fusion Framework},
author={Ying, Zhenqiang and Li, Ge and Ren, Yurui and Wang, Ronggang and Wang, Wenmin},
booktitle={International Conference on Computer Analysis of Images and Patterns},
pages={36--46},
year={2017},
organization={Springer}
}

First Online: 28 July 2017 Springer

Matlab Implementation:

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
function fused = Ying_2017_CAIP(I, mu, k, a, b) % camera a, b
if ~exist( 'mu', 'var' )
mu = 0.5;
end
if ~exist( 'a', 'var' )
a = -0.3293;
end
if ~exist( 'b', 'var' )
b = 1.1258;
end
if ~isfloat(I)
I = im2double( I );
end
lambda = 0.5;
sigma = 5;
%% t: scene illumination map
t_b = max( I, [], 3 ); % also work for single-channel image
t_our = imresize( tsmooth( imresize( t_b, 0.5 ), lambda, sigma ), size( t_b ) );
%% k: exposure ratio
if ~exist( 'k', 'var' ) || isempty(k)
isBad = t_our < 0.5;
J = maxEntropyEnhance(I, isBad);
else
J = applyK(I, k, a, b); %k
J = min(J, 1); % fix overflow
end
%% W: Weight Matrix
t = repmat(t_our, [1 1 size(I,3)]);
W = t.^mu;
I2 = I.*W;
J2 = J.*(1-W);
fused = I2 + J2;
function J = maxEntropyEnhance(I, isBad)
Y = rgb2gm(real(max(imresize(I, [50 50]), 0))); % max - avoid complex number
if exist('isBad', 'var')
isBad = (imresize(isBad, [50 50]));
Y = Y(isBad);
end
if isempty(Y)
J = I; % no enhancement k = 1
return;
end
opt_k = fminbnd(@(k) ( -entropy(applyK(Y, k)) ),1, 7);
J = applyK(I, opt_k, a, b) - 0.01;
end
end
function I = rgb2gm(I)
if size(I,3) == 3
I = im2double(max(0,I)); % negative double --> complex double
I = ( I(:,:,1).*I(:,:,2).*I(:,:,3) ).^(1/3);
end
end
function J = applyK(I, k, a, b)
if ~exist( 'a', 'var' )
a = -0.3293;
end
if ~exist( 'b', 'var' )
b = 1.1258;
end
f = @(x)exp((1-x.^a)*b);
beta = f(k);
gamma = k.^a;
J = I.^gamma.*beta;
end
function S = tsmooth( I, lambda, sigma, sharpness)
if ( ~exist( 'lambda', 'var' ) )
lambda = 0.01;
end
if ( ~exist( 'sigma', 'var' ) )
sigma = 3.0;
end
if ( ~exist( 'sharpness', 'var' ) )
sharpness = 0.001;
end
I = im2double( I );
x = I;
[ wx, wy ] = computeTextureWeights( x, sigma, sharpness);
S = solveLinearEquation( I, wx, wy, lambda );
end
function [ W_h, W_v ] = computeTextureWeights( fin, sigma, sharpness)
dt0_v = [diff(fin,1,1);fin(1,:)-fin(end,:)];
dt0_h = [diff(fin,1,2)';fin(:,1)'-fin(:,end)']';
gauker_h = filter2(ones(1,sigma),dt0_h);
gauker_v = filter2(ones(sigma,1),dt0_v);
W_h = 1./(abs(gauker_h).*abs(dt0_h)+sharpness);
W_v = 1./(abs(gauker_v).*abs(dt0_v)+sharpness);
end
function OUT = solveLinearEquation( IN, wx, wy, lambda )
[ r, c, ch ] = size( IN );
k = r * c;
dx = -lambda * wx( : );
dy = -lambda * wy( : );
tempx = [wx(:,end),wx(:,1:end-1)];
tempy = [wy(end,:);wy(1:end-1,:)];
dxa = -lambda *tempx(:);
dya = -lambda *tempy(:);
tempx = [wx(:,end),zeros(r,c-1)];
tempy = [wy(end,:);zeros(r-1,c)];
dxd1 = -lambda * tempx(:);
dyd1 = -lambda * tempy(:);
wx(:,end) = 0;
wy(end,:) = 0;
dxd2 = -lambda * wx(:);
dyd2 = -lambda * wy(:);
Ax = spdiags( [dxd1,dxd2], [-k+r,-r], k, k );
Ay = spdiags( [dyd1,dyd2], [-r+1,-1], k, k );
D = 1 - ( dx + dy + dxa + dya);
A = (Ax+Ay) + (Ax+Ay)' + spdiags( D, 0, k, k );
if exist( 'ichol', 'builtin' )
L = ichol( A, struct( 'michol', 'on' ) );
OUT = IN;
for ii = 1:ch
tin = IN( :, :, ii );
[ tout, ~ ] = pcg( A, tin( : ), 0.1, 50, L, L' );
OUT( :, :, ii ) = reshape( tout, r, c );
end
else
OUT = IN;
for ii = 1:ch
tin = IN( :, :, ii );
tout = A\tin( : );
OUT( :, :, ii ) = reshape( tout, r, c );
end
end
end

Try it:

1
2
3
4
I = imread('yellowlily.jpg');
J = Ying_2017_CAIP(I);
subplot 121; imshow(I); title('Original Image');
subplot 122; imshow(J); title('Enhanced Result');