資源簡介
這里主要講深度學習用在超分辨率重建上的開山之作SRCNN。超分辨率技術(Super-Resolution)是指從觀測到的低分辨率圖像重建出相應的高分辨率圖像,在監控設備、衛星圖像和醫學影像等領域都有重要的應用價值。SR可分為兩類:從多張低分辨率圖像重建出高分辨率圖像和從單張低分辨率圖像重建出高分辨率圖像。基于深度學習的SR,主要是基于單張低分辨率的重建方法,即Single Image Super-Resolution (SISR)。
SR方法主要可以分為四種模型:基于邊緣,基于圖像統計,基于樣本(基于補丁)的方法。本文的SRCNN網絡結構非常簡單,僅僅只有三層網絡就是實現了SR。網絡結構如下圖所示:
代碼片段和文件信息
clear;close?all;
%%?settings
folder?=?‘Test/Set5‘;
savepath?=?‘test.h5‘;
size_input?=?33;
size_label?=?21;
scale?=?3;
stride?=?21?;
%%?initialization
data?=?zeros(size_input?size_input?1?1);
label?=?zeros(size_label?size_label?1?1);
padding?=?abs(size_input?-?size_label)/2;
count?=?0;
%%?generate?data
filepaths?=?dir(fullfile(folder‘*.bmp‘));
????
for?i?=?1?:?length(filepaths)
????
????image?=?imread(fullfile(folderfilepaths(i).name));
????image?=?rgb2ycbcr(image);
????image?=?im2double(image(:?:?1));
????
????im_label?=?modcrop(image?scale);
????[heiwid]?=?size(im_label);
????im_input?=?imresize(imresize(im_label1/scale‘bicubic‘)[heiwid]‘bicubic‘);
????for?x?=?1?:?stride?:?hei-size_input+1
????????for?y?=?1?:stride?:?wid-size_input+1
????????????
????????????subim_input?=?im_input(x?:?x+size_input-1?y?:?y+size_input-1);
????????????subim_label?=?im_label(x+padding?:?x+padding+size_label-1?y+padding?:?y+padding+size_label-1);
????????????count=count+1;
????????????data(:?:?1?count)?=?subim_input;
????????????label(:?:?1?count)?=?subim_label;
????????end
????end
end
order?=?randperm(count);
data?=?data(:?:?1?order);
label?=?label(:?:?1?order);?
%%?writing?to?HDF5
chunksz?=?2;
created_flag?=?false;
totalct?=?0;
for?batchno?=?1:floor(count/chunksz)
????last_read=(batchno-1)*chunksz;
????batchdata?=?data(::1last_read+1:last_read+chunksz);?
????batchlabs?=?label(::1last_read+1:last_read+chunksz);
????startloc?=?struct(‘dat‘[111totalct+1]?‘lab‘?[111totalct+1]);
????curr_dat_sz?=?store2hdf5(savepath?batchdata?batchlabs?~created_flag?startloc?chunksz);?
????created_flag?=?true;
????totalct?=?curr_dat_sz(end);
end
h5disp(savepath);
?屬性????????????大小?????日期????時間???名稱
-----------?---------??----------?-----??----
?????文件???????1771??2015-07-08?11:32??SRCNN\generate_test.m
?????文件???????1769??2015-07-08?11:31??SRCNN\generate_train.m
?????文件????????279??2015-03-17?01:58??SRCNN\modcrop.m
?????文件???????1088??2015-07-01?12:36??SRCNN\Readme.txt
?????文件???????1318??2015-07-01?12:32??SRCNN\saveFilters.m
?????文件???????1478??2015-07-01?11:25??SRCNN\SRCNN_mat.prototxt
?????文件???????1705??2015-07-01?12:27??SRCNN\SRCNN_net.prototxt
?????文件????????566??2015-07-01?11:29??SRCNN\SRCNN_solver.prototxt
?????文件???????2897??2015-06-05?03:14??SRCNN\store2hdf5.m
?????文件?????720054??2015-03-17?01:58??SRCNN\Test\Set14\baboon.bmp
?????文件????1244214??2015-03-17?01:58??SRCNN\Test\Set14\barbara.bmp
?????文件?????263222??2015-03-17?01:58??SRCNN\Test\Set14\bridge.bmp
?????文件?????304182??2015-03-17?01:58??SRCNN\Test\Set14\coastguard.bmp
?????文件?????271526??2015-03-17?01:58??SRCNN\Test\Set14\comic.bmp
?????文件?????228584??2015-03-17?01:58??SRCNN\Test\Set14\face.bmp
?????文件?????543054??2015-03-17?01:58??SRCNN\Test\Set14\flowers.bmp
?????文件?????304182??2015-03-17?01:58??SRCNN\Test\Set14\foreman.bmp
?????文件?????786486??2015-03-17?01:58??SRCNN\Test\Set14\lenna.bmp
?????文件?????786486??2015-03-17?01:58??SRCNN\Test\Set14\man.bmp
?????文件????1179702??2015-03-17?01:58??SRCNN\Test\Set14\monarch.bmp
?????文件?????786486??2015-03-17?01:58??SRCNN\Test\Set14\pepper.bmp
?????文件????1041782??2015-03-17?01:58??SRCNN\Test\Set14\ppt3.bmp
?????文件?????688214??2015-03-17?01:58??SRCNN\Test\Set14\zebra.bmp
?????文件?????786486??2015-03-17?01:58??SRCNN\Test\Set5\baby_GT.bmp
?????文件?????248886??2015-03-17?01:58??SRCNN\Test\Set5\bird_GT.bmp
?????文件?????196730??2015-03-17?01:58??SRCNN\Test\Set5\butterfly_GT.bmp
?????文件?????235254??2015-03-17?01:58??SRCNN\Test\Set5\head_GT.bmp
?????文件?????235350??2015-03-17?01:58??SRCNN\Test\Set5\woman_GT.bmp
?????文件????6887544??2018-02-17?22:20??SRCNN\test.h5
?????文件?????????23??2015-07-01?11:24??SRCNN\test.txt
............此處省略101個文件信息
評論
共有 條評論