-
Notifications
You must be signed in to change notification settings - Fork 275
/
sr_quant_eval.m
125 lines (110 loc) · 5.37 KB
/
sr_quant_eval.m
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
% SR_QUANT_EVAL
%
% Script for computing averaged PSNR, SSIM, and IVC.
% The script reproduces the quantitative results on our project page on
% Github: https://github.com/jbhuang0604/SelfExSR
%
% Jia-Bin Huang
% Electrical and Computer Engineering
% University of Illinois, Urbana-Champaign
% www.jiabinhuang.com
clc;
clear;
close all;
% Add pathes for running SSIM and IVC
addpath(genpath('quant_eval'));
% Dataset list
numDataset = 4;
dataset = cell(numDataset,1);
dataset{1}.name = 'Urban100';
dataset{2}.name = 'BSD100';
dataset{3}.name = 'Set5';
dataset{4}.name = 'Set14';
% Super-resolution factor
dataset{1}.SRF = [2, 4];
dataset{2}.SRF = [2, 3, 4];
dataset{3}.SRF = [2, 3, 4];
dataset{4}.SRF = [2, 3, 4];
% Number of images in the dataset
dataset{1}.numImg = 100;
dataset{2}.numImg = 100;
dataset{3}.numImg = 5;
dataset{4}.numImg = 14;
% Method list
methodList = {'Bicubic', 'ScSR', 'Kim', 'Abhishek', 'Glasner', 'SRCNN', 'A+', 'SelfExSR'};
numMethod = length(methodList);
indValidMethod = [1, 1, 1, 1, 1, 1, 1, 1]; % Test only
% Initialize result path
resPath = fullfile('quant_eval', 'result');
if(~exist(resPath, 'dir'))
mkdir(resPath);
end
% =========================================================================
% Start quantitative evaluation
% =========================================================================
for indDataset = 1 : numDataset
datasetName = dataset{indDataset}.name;
numImgDataset = dataset{indDataset}.numImg;
numSRF = size(dataset{indDataset}.SRF,2);
% Run each super-resolution factor
for indSRF = 1: numSRF
SRF = dataset{indDataset}.SRF(indSRF);
imgPath = fullfile('data', datasetName, ['image_SRF_',num2str(SRF)]);
resName = ['result_', datasetName, '_SRF_', num2str(SRF),'.mat'];
if(~exist(fullfile(resPath, resName), 'file'))
% Initialize result table
SSIM_table = zeros(numImgDataset, numMethod);
PSNR_table = zeros(numImgDataset, numMethod);
IFC_table = zeros(numImgDataset, numMethod);
reverseStr = '';
for indImg = 1: numImgDataset
% Load groundtruth high-resolution image
imgName = ['img_',num2str(indImg, '%03d'), '_SRF_', num2str(SRF), '_HR.png'];
imgGT = imread(fullfile(imgPath, imgName));
% Load super-resolution results
for indMethod= 1:numMethod
if(indValidMethod(indMethod))
imgName = ['img_',num2str(indImg, '%03d'), '_SRF_', num2str(SRF), '_', methodList{indMethod}, '.png'];
img = imread(fullfile(imgPath, imgName), 'png');
% Compute image quality
[psnr, ssim, ifc] = compute_difference(img, imgGT, SRF);
PSNR_table(indImg,indMethod) = psnr;
SSIM_table(indImg,indMethod) = ssim;
IFC_table(indImg, indMethod) = ifc;
end
% Display progress
percentDone = 100*((indImg-1)*numMethod + indMethod)/(numMethod*numImgDataset);
msg = sprintf('Evaluating dataset %s on SRF %d, progress: %3.2f', datasetName, SRF, percentDone);
fprintf([reverseStr, msg]);
reverseStr = repmat(sprintf('\b'), 1, length(msg));
end
end
% Save results
resName = ['result_', datasetName, '_SRF_', num2str(SRF),'.mat'];
save(fullfile(resPath, resName), 'PSNR_table', 'SSIM_table', 'IFC_table');
else
load(fullfile(resPath, resName));
end
% Display results
avgPSNR = mean(PSNR_table,1);
avgSSIM = mean(SSIM_table,1);
avgIFC = mean(IFC_table, 1);
fprintf('\n\n=== Quantitative results for dataset %s on SRF %d === \n\n', datasetName, SRF);
fprintf('Peak signal-to-noise ratio (PSNR) \n')
fprintf(' %8s\t%8s\t%8s\t%8s\t%8s\t%8s\t%8s\t%8s\t \n', methodList{1}, methodList{2}, ...
methodList{3}, methodList{4}, methodList{5}, methodList{6}, methodList{7}, methodList{8});
fprintf('PSNR|%8.02f\t|%8.02f\t|%8.02f\t|%8.02f\t|%8.02f\t|%8.02f\t|%8.02f\t|%8.02f\t| \n', ...
avgPSNR(1), avgPSNR(2), avgPSNR(3), avgPSNR(4), avgPSNR(5), avgPSNR(6), avgPSNR(7), avgPSNR(8));
fprintf('Structural similarity index (SSIM) \n')
fprintf(' %8s\t%8s\t%8s\t%8s\t%8s\t%8s\t%8s\t%8s\t \n', methodList{1}, methodList{2}, ...
methodList{3}, methodList{4}, methodList{5}, methodList{6}, methodList{7}, methodList{8});
fprintf('SSIM|%8.04f\t|%8.04f\t|%8.04f\t|%8.04f\t|%8.04f\t|%8.04f\t|%8.04f\t|%8.04f\t| \n', ...
avgSSIM(1), avgSSIM(2), avgSSIM(3), avgSSIM(4), avgSSIM(5), avgSSIM(6), avgSSIM(7), avgSSIM(8));
fprintf('Information fidelity criterion (IFC) \n')
fprintf(' %8s\t%8s\t%8s\t%8s\t%8s\t%8s\t%8s\t%8s\t \n', methodList{1}, methodList{2}, ...
methodList{3}, methodList{4}, methodList{5}, methodList{6}, methodList{7}, methodList{8});
fprintf('IFC|%8.02f\t|%8.02f\t|%8.02f\t|%8.02f\t|%8.02f\t|%8.02f\t|%8.02f\t|%8.02f\t| \n\n', ...
avgIFC(1), avgIFC(2), avgIFC(3), avgIFC(4), avgIFC(5), avgIFC(6), avgIFC(7), avgIFC(8));
fprintf('\n');
end
end