{"id":5249,"date":"2022-08-07T21:33:25","date_gmt":"2022-08-07T13:33:25","guid":{"rendered":"http:\/\/139.9.1.231\/?p=5249"},"modified":"2022-08-07T21:33:26","modified_gmt":"2022-08-07T13:33:26","slug":"real-time-single-image-and-video-super-resolution-using-an-efficient-sub-pixel-convolutional-neural-network","status":"publish","type":"post","link":"http:\/\/139.9.1.231\/index.php\/2022\/08\/07\/real-time-single-image-and-video-super-resolution-using-an-efficient-sub-pixel-convolutional-neural-network\/","title":{"rendered":"Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network"},"content":{"rendered":"\n

\u8bba\u6587\u6807\u9898\uff1aReal-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network<\/a><\/p>\n\n\n\n

github\uff1a https:\/\/github.com\/leftthomas\/ESPCN<\/a><\/p>\n\n\n\n

2016\u5e74\u7684\u6587\u7ae0\u3002\u5728\u6b64\u4e4b\u524d\u4f7f\u7528CNN\u8fdb\u884cSR\u7684\u65b9\u6cd5\u90fd\u662f\u5c06LR\u56fe\u50cf\u5148\u7528\u4e00\u4e2asingle filter\uff08\u901a\u5e38\u662fbicubic\uff09upscale\u81f3HR\u7684\u5c3a\u5bf8\uff0c\u518d\u8fdb\u884creconstruction\u7684\u3002\u6240\u6709SR\u7684\u64cd\u4f5c\u90fd\u518dHR\u7a7a\u95f4\u8fdb\u884c\u3002\u3000\u3000\u800c\u672c\u6587\u63d0\u51fa\u5728LR\u7a7a\u95f4\u8fdb\u884c\u7279\u5f81\u63d0\u53d6\u3002\u5e76\u5f15\u5165sub-pixel convolution layer\u7528\u4e8e\u5b66\u4e60\u4e00\u7ec4upscaling filter\uff0c\u7528\u8fd9\u4e9b\u9488\u5bf9\u7279\u5f81\u56fe\u8bad\u7ec3\u5f97\u5230\u7684\u66f4\u590d\u6742\u7684filter\u4ee3\u66ff\u624b\u5de5bicubic filter\u3002\u53ef\u4ee5\u964d\u4f4e\u8ba1\u7b97\u6210\u672c\uff0c\u5b9e\u73b0\u5b9e\u65f6SR\u3002\u76f4\u63a5\u5c06LR\u56fe\u50cf\u8f93\u5165\u4e00\u4e2al\u5c42\u7684CNN\u4e2d\uff0c\u4e4b\u540e\u901a\u8fc7\u4e00\u5c42sub-pixel\u5377\u79ef\u5c42upscaleLR\u7279\u5f81\u56fe\u751f\u6210\u5bf9\u5e94\u7684HR\u56fe\u50cf\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u8fd9\u7bc7\u8bba\u6587\u63d0\u51fa\u4e86\u4e00\u79cd\u4e9a\u50cf\u7d20\u5377\u79ef\u7684\u65b9\u6cd5\u6765\u5bf9\u56fe\u50cf\u8fdb\u884c\u8d85\u5206\u8fa8\u7387\u91cd\u5efa\uff0c\u901f\u5ea6\u7279\u522b\u5feb\u3002\u867d\u7136\u8bba\u6587\u91cc\u9762\u79f0\u63d0\u51fa\u7684\u65b9\u6cd5\u4e3a\u4e9a\u50cf\u7d20\u5377\u79ef(sub-pixel convolution)\uff0c\u4f46\u662f\u5b9e\u9645\u4e0a\u5e76\u4e0d\u6d89\u53ca\u5230\u5377\u79ef\u8fd0\u7b97\uff0c\u662f\u4e00\u79cd\u9ad8\u6548\u3001\u5feb\u901f\u3001\u65e0\u53c2\u7684\u50cf\u7d20\u91cd\u6392\u5217\u7684\u4e0a\u91c7\u6837\u65b9\u5f0f\u3002\u7531\u4e8e\u5f88\u5feb\uff0c\u76f4\u63a5\u7528\u5728\u89c6\u9891\u8d85\u5206\u4e2d\u4e5f\u53ef\u4ee5\u505a\u5230\u5b9e\u65f6\u3002\u5176\u5728Tensorflow\u4e2d\u7684\u5b9e\u73b0\u79f0\u4e3adepthto<\/em>space<\/strong>\u00a0,\u5728Pytorch\u4e2d\u7684\u5b9e\u73b0\u4e3aPixelShuffle<\/strong>\u3002<\/p>\n\n\n\n

\u8fd9\u79cd\u4e0a\u91c7\u6837\u7684\u65b9\u5f0f\u5f88\u591a\u65f6\u5019\u90fd\u6210\u4e3a\u4e86\u4e0a\u91c7\u6837\u7684\u9996\u9009\uff0c\u7ecf\u5e38\u7528\u5728\u56fe\u50cf\u91cd\u5efa\u9886\u57df\uff0c\u5982\u540e\u7eed\u6709\u5728\u964d\u566a\u9886\u57df\u4e2d\u7684FFDNet<\/a>\u3002<\/p>\n\n\n\n

\u8bba\u6587\u7684\u4e3b\u8981\u521b\u65b0\u70b9\u4e3a\uff1a<\/p>\n\n\n\n

1. \u53ea\u5728\u6a21\u578b\u672b\u7aef\u8fdb\u884c\u4e0a\u91c7\u6837\uff0c\u53ef\u4ee5\u4f7f\u5f97\u5728\u4f4e\u5206\u8fa8\u7387\u7a7a\u95f4\u4fdd\u7559\u66f4\u591a\u7684\u7eb9\u7406\u533a\u57df\uff0c\u5728\u89c6\u9891\u8d85\u5206\u4e2d\u4e5f\u53ef\u4ee5\u505a\u5230\u5b9e\u65f6\u3002
2.\u6a21\u5757\u672b\u7aef\u76f4\u63a5\u4f7f\u7528\u4e9a\u50cf\u7d20\u5377\u79ef\u7684\u65b9\u5f0f\u6765\u8fdb\u884c\u4e0a\u91c7\u6837\uff0c\u76f8\u6bd4\u4e8e\u663e\u793a\u7684\u5c06LR\u63d2\u503c\u5230HR\uff0c\u8fd9\u79cd\u4e0a\u91c7\u6837\u65b9\u5f0f\u53ef\u4ee5\u5b66\u4e60\u5230\u66f4\u597d\u3001\u66f4\u4e3a\u590d\u6742\u7684\u65b9\u5f0f\uff0c\u53ef\u4ee5\u83b7\u5f97\u66f4\u597d\u7684\u91cd\u5efa\u6548\u679c\u3002<\/p><\/blockquote>\n\n\n\n

\u53ef\u4ee5\u770b\u5230\uff0c\u76f8\u6bd4\u4e8e\u5176\u4ed6\u7684\u4e00\u4e9b\u8d85\u5206\u7b97\u6cd5\uff0c\u8fd9\u91cc\u5b9e\u9645\u4e0a\u53ea\u6539\u52a8\u4e86\u6700\u540e\u7684\u4e0a\u91c7\u6837\u65b9\u5f0f\u3002\u5728\u6a21\u578b\u5012\u6570\u7b2c\u4e8c\u5c42\u5b66\u4e60\u5bf9\u5e94\u7684\u901a\u9053\u6570( r2c )\u7684\u5377\u79ef\uff0c\u5176\u4e2dc\u4e3a\u6700\u7ec8\u7684\u901a\u9053\u6570\uff0c\u5982\u8f93\u51fa\u662fRGB\u56fe\uff0c\u5219c\u4e3a3\uff0c\u5982\u8f93\u51fa\u662f\u7070\u5ea6\u56fe\u6216\u8005Y\u901a\u9053\u7684\u56fe\uff0c\u5219c\u4e3a1\uff1br\u4e3a\u9700\u8981\u8fdb\u884c\u7684\u4e0a\u91c7\u6837\u500d\u6570\uff0c\u4e3a\u6b63\u6574\u6570\u500d\uff0c\u4e0d\u540c\u7684\u4e0a\u91c7\u6837\u500d\u6570\u53ea\u9700\u8981\u8c03\u6574\u8fd9\u4e00\u5c42\u5377\u79ef\u7684\u901a\u9053\u6570\u5373\u53ef\u3002<\/p>\n\n\n\n

\u7531\u4e8e\u5e26\u8ba1\u7b97\u7684\u64cd\u4f5c\u90fd\u662f\u5728\u4f4e\u5206\u8fa8\u7387\u7a7a\u95f4\u4e2d\u8fdb\u884c\u7684\uff0c\u6240\u4ee5\u901f\u5ea6\u76f8\u5bf9\u4f1a\u5feb\u5f88\u591a\u3002<\/p>\n\n\n\n

\u8fd9\u91cc\u7ed9\u51fa\u7684\u793a\u4f8b\u4e3ar=3,c=1\uff0c\u5373\u5355\u901a\u9053\u56fe\u76843\u500d\u4e0a\u91c7\u6837\u56fe\u3002\u7ed3\u5408\u8d85\u50cf\u7d20\u7684\u601d\u60f3\u6765\u770b\uff0c\u4ee5\u7b2c\u4e00\u5f20\u7279\u5f81\u56fe\u8fdb\u884c\u76843×3\u5bab\u683c\u7684\u50cf\u7d20\u91cd\u6392\u5217\uff0c\u884c\u4f18\u5148\u5730\u6309\u987a\u5e8f\u5c06\u4e0d\u540c\u6df1\u5ea6\u7684\u7279\u5f81\u4f9d\u6b21\u91cd\u6392\u5217\u5230\u5bab\u683c\u4e0a\u3002<\/p>\n\n\n\n

\u8fd9\u91cc\u7ed9\u51fa\u7684\u793a\u4f8b\u56fe\u662f\u7b80\u5355\u573a\u666f\uff0c\u50cf\u7d20\u91cd\u6392\u5217\u7684\u65b9\u5f0f\u4e3a\uff1a<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u4ece\u516c\u5f0f\u53ef\u4ee5\u770b\u51fa\uff0c\u5bf9\u4e8e\u591a\u901a\u9053\u7684\u56fe\uff0c\u4ee5\u901a\u9053\u6570\u4f5c\u4e3a\u4e00\u4e2a\u6574\u4f53\uff0c\u5373\u5c06\u7279\u5f81\u56fe\u901a\u9053\u6570\u4e2d\u8fde\u7eed\u7684c\u4e2a\u901a\u9053\u4f5c\u4e3a\u4e00\u4e2a\u6574\u4f53\uff0c\u518d\u7136\u540e\u8fdb\u884c\u50cf\u7d20\u91cd\u6392\u5217\uff0c\u5f97\u5230\u591a\u901a\u9053\u7684\u4e0a\u91c7\u6837\u56fe\u3002<\/p>\n\n\n\n

\u8bba\u6587\u7684\u6838\u5fc3\u521b\u65b0\u70b9\u5c31\u5728\u4e8e\u8fd9\u91cc\u7684\u50cf\u7d20\u91cd\u6392\u5217\u7684\u65b9\u5f0f\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u6574\u4f53\u7684\u6548\u679c\u4e0a\u6765\u8bf4\uff0c\u4e5f\u662f\u975e\u5e38\u7684\u60ca\u4eba\u3002\u4ece\u6a21\u578b\u7684\u89d2\u5ea6\u4e0a\u800c\u8a00\uff0c\u5176\u4e3b\u5e72\u6a21\u578b\u53ef\u4ee5\u91c7\u6837\u5176\u4ed6SOTA\u7684\u4e3b\u5e72\u7ed3\u6784\u3002\u7531\u4e8e\u4e0a\u91c7\u6837\u7684\u5dee\u5f02\uff0c\u53ef\u4ee5\u5b66\u4e60\u5230\u66f4\u597d\u3001\u66f4\u590d\u6742\u7684\u4e0a\u91c7\u6837\u65b9\u5f0f\uff0c\u6240\u4ee5\u6700\u7ec8\u7684\u91cd\u5efa\u6548\u679c\u662f\u8981\u7a0d\u597d\u4e8eSOTA\u7684\u6a21\u578b\u7684\u3002\u5e76\u4e14\u7531\u4e8e\u4e0a\u91c7\u6837\u7279\u522b\u9ad8\u6548\uff0c\u901f\u5ea6\u975e\u5e38\u7684\u5feb\u3002\u4ecePSNR\u7684\u89d2\u5ea6\u6765\u770b\uff0cESPCN\u6bd4TNRD(TNRD\u53d1\u8868\u4e8eTPAMI2015\uff0c\u662fDnCNN<\/a>\u7684\u524d\u8eab)\u7a0d\u597d\uff0c\u4f46\u662f\u901f\u5ea6\u5374\u76f8\u5dee\u4e00\u4e2a\u6570\u91cf\u7ea7\u5de6\u53f3\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u7ed3\u8bba<\/strong><\/h2>\n\n\n\n

\u8fd9\u7bc7\u8bba\u6587\u63d0\u51fa\u4e86\u4e00\u79cd\u4e9a\u50cf\u7d20\u5377\u79ef\u5c42\uff0c\u5728\u4f4e\u5206\u8fa8\u7387\u7a7a\u95f4\u4e2d\u53ef\u4ee5\u5b66\u4e60\u5230\u66f4\u597d\u3001\u66f4\u590d\u6742\u7684\u4e0a\u91c7\u6837\u65b9\u5f0f\uff0c\u5bf9\u4e8e\u4e0d\u540c\u7684\u91cd\u5efa\u500d\u6570\uff0c\u53ea\u9700\u8981\u5bf9\u5e94\u5730\u66f4\u6539\u4f4e\u5206\u8fa8\u7387\u7a7a\u95f4\u4e2d\u7684\u5377\u79ef\u901a\u9053\u6570\uff0c\u975e\u5e38\u7075\u6d3b\u3002\u5176\u6700\u7ec8\u7684\u91cd\u5efaPSNR\u6548\u679c\u4e5f\u662fSOTA\uff0c\u901f\u5ea6\u4e0a\u5177\u6709\u5f88\u5de8\u5927\u7684\u4f18\u52bf\uff0c\u89c6\u9891\u8d85\u5206\u4e5f\u80fd\u505a\u5230\u5b9e\u65f6\u3002\u8fd9\u79cd\u4e0a\u91c7\u6837\u65b9\u5f0f\u4e5f\u5e7f\u6cdb\u5730\u5e94\u7528\u4e8e\u5176\u4ed6\u7684\u91cd\u5efa\u9886\u57df\u4e2d\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

\u8bba\u6587\u6807\u9898\uff1aReal-Time Single Image and Video Super-Resolution … \u7ee7\u7eed\u9605\u8bfbReal-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[18],"tags":[],"_links":{"self":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts\/5249"}],"collection":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/comments?post=5249"}],"version-history":[{"count":7,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts\/5249\/revisions"}],"predecessor-version":[{"id":5260,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts\/5249\/revisions\/5260"}],"wp:attachment":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/media?parent=5249"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/categories?post=5249"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/tags?post=5249"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}