SRcnn:神经网络重建图片的开山之作

% =========================================================================
% Test code for Super-Resolution Convolutional Neural Networks (SRCNN)
%
% Reference
%   Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang. Learning a Deep Convolutional Network for Image Super-Resolution, 
%   in Proceedings of European Conference on Computer Vision (ECCV), 2014
%
%   Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang. Image Super-Resolution Using Deep Convolutional Networks,
%   arXiv:1501.00092
%
% Chao Dong
% IE Department, The Chinese University of Hong Kong
% For any question, send email to ndc.forward@gmail.com
% =========================================================================

close all;
clear all;

%% read ground truth image
im  = imread('Set5\butterfly_GT.bmp');
%im  = imread('Set14\zebra.bmp');

%% set parameters
up_scale = 3;
model = 'model\9-5-5(ImageNet)\x3.mat';
% up_scale = 3;
% model = 'model\9-3-5(ImageNet)\x3.mat';
% up_scale = 3;
% model = 'model\9-1-5(91 images)\x3.mat';
% up_scale = 2;
% model = 'model\9-5-5(ImageNet)\x2.mat'; 
% up_scale = 4;
% model = 'model\9-5-5(ImageNet)\x4.mat';

%% work on illuminance only
if size(im,3)>1
    im = rgb2ycbcr(im);
    im = im(:, :, 1);
end
im_gnd = modcrop(im, up_scale);   %保证图像被up_scale整除
im_gnd = single(im_gnd)/255;     %Single（单精度浮点型）变量存储为 IEEE 32 位（4 个字节）浮点数值的形式，它的范围在负数的时候是从 -3.402823E38 到 -1.401298E-45，而在正数的时候是从 1.401298E-45 到 3.402823E38。

%% bicubic interpolation    
im_l = imresize(im_gnd, 1/up_scale, 'bicubic');    %缩小3倍
im_b = imresize(im_l, up_scale, 'bicubic');        %又放大三倍

%% SRCNN
im_h = SRCNN(model, im_b);               %用网络处理一下

%% remove border  %去除没有用的边界
im_h = shave(uint8(im_h * 255), [up_scale, up_scale]);  %表示变量是无符号整数，范围是0到255.
im_gnd = shave(uint8(im_gnd * 255), [up_scale, up_scale]);
im_b = shave(uint8(im_b * 255), [up_scale, up_scale]);

%% compute PSNR
psnr_bic = compute_psnr(im_gnd,im_b);
psnr_srcnn = compute_psnr(im_gnd,im_h);

%% show results
fprintf('PSNR for Bicubic Interpolation: %f dB\n', psnr_bic);
fprintf('PSNR for SRCNN Reconstruction: %f dB\n', psnr_srcnn);

%保存 图片
imwrite(im_h,'img_h.png');
imwrite(im_b,'img_b.png');
imwrite(im_gnd,'img_gnd.png');



figure, imshow(im_b); title('Bicubic Interpolation');
figure, imshow(im_h); title('SRCNN Reconstruction');

%imwrite(im_b, ['Bicubic Interpolation' '.bmp']);
%imwrite(im_h, ['SRCNN Reconstruction' '.bmp']);

SRCNN的核心算法：

function im_h = SRCNN(model, im_b)

%% load CNN model parameters
load(model);
[conv1_patchsize2,conv1_filters] = size(weights_conv1);
conv1_patchsize = sqrt(conv1_patchsize2);
[conv2_channels,conv2_patchsize2,conv2_filters] = size(weights_conv2);
conv2_patchsize = sqrt(conv2_patchsize2);
[conv3_channels,conv3_patchsize2] = size(weights_conv3);
conv3_patchsize = sqrt(conv3_patchsize2);
[hei, wid] = size(im_b);

%% conv1
weights_conv1 = reshape(weights_conv1, conv1_patchsize, conv1_patchsize, conv1_filters);
conv1_data = zeros(hei, wid, conv1_filters);
for i = 1 : conv1_filters
    conv1_data(:,:,i) = imfilter(im_b, weights_conv1(:,:,i), 'same', 'replicate');
    conv1_data(:,:,i) = max(conv1_data(:,:,i) + biases_conv1(i), 0);
end

%% conv2
conv2_data = zeros(hei, wid, conv2_filters);
for i = 1 : conv2_filters
    for j = 1 : conv2_channels
        conv2_subfilter = reshape(weights_conv2(j,:,i), conv2_patchsize, conv2_patchsize);
        conv2_data(:,:,i) = conv2_data(:,:,i) + imfilter(conv1_data(:,:,j), conv2_subfilter, 'same', 'replicate');
    end
    conv2_data(:,:,i) = max(conv2_data(:,:,i) + biases_conv2(i), 0);
end

%% conv3
conv3_data = zeros(hei, wid);
for i = 1 : conv3_channels
    conv3_subfilter = reshape(weights_conv3(i,:), conv3_patchsize, conv3_patchsize);
    conv3_data(:,:) = conv3_data(:,:) + imfilter(conv2_data(:,:,i), conv3_subfilter, 'same', 'replicate');
end

%% SRCNN reconstruction
im_h = conv3_data(:,:) + biases_conv3;

图解里面变量和卷积