使用反向传播神经网络的单幅图像超分辨率

2017 20th International Conference of Computer and Information Technology (ICCIT) Pub Date : 2017-12-22 DOI:10.1109/ICCITECHN.2017.8281778

M. S. Hasan, Salman Taseen Haque

{"title":"使用反向传播神经网络的单幅图像超分辨率","authors":"M. S. Hasan, Salman Taseen Haque","doi":"10.1109/ICCITECHN.2017.8281778","DOIUrl":null,"url":null,"abstract":"There are several existing mathematical algorithms for colour image upscaling like Nearest Neighbour, Bicubic and Bilinear. This paper firstly investigates the performances of these three and it has been found that Bicubic performs the best in terms of structural similarity and execution time. A Bicubic with backpropagation based ANN method has been proposed to improve the results. Bicubic with ANN shows 6.5% higher SSIM, 6.9% higher PSNR, 8.7% higher SNR and 30.23% lower MSE values than pure Bicubic. The results of Bicubic with ANN are also compared with state of the art super-resolution techniques like SRCNN. Bicubic with ANN produces 1.48% higher SSIM and 3.44% higher PSNR than SRCNN.","PeriodicalId":350374,"journal":{"name":"2017 20th International Conference of Computer and Information Technology (ICCIT)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2017-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"Single image super-resolution using back-propagation neural networks\",\"authors\":\"M. S. Hasan, Salman Taseen Haque\",\"doi\":\"10.1109/ICCITECHN.2017.8281778\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"There are several existing mathematical algorithms for colour image upscaling like Nearest Neighbour, Bicubic and Bilinear. This paper firstly investigates the performances of these three and it has been found that Bicubic performs the best in terms of structural similarity and execution time. A Bicubic with backpropagation based ANN method has been proposed to improve the results. Bicubic with ANN shows 6.5% higher SSIM, 6.9% higher PSNR, 8.7% higher SNR and 30.23% lower MSE values than pure Bicubic. The results of Bicubic with ANN are also compared with state of the art super-resolution techniques like SRCNN. Bicubic with ANN produces 1.48% higher SSIM and 3.44% higher PSNR than SRCNN.\",\"PeriodicalId\":350374,\"journal\":{\"name\":\"2017 20th International Conference of Computer and Information Technology (ICCIT)\",\"volume\":\"39 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2017-12-22\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2017 20th International Conference of Computer and Information Technology (ICCIT)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICCITECHN.2017.8281778\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2017 20th International Conference of Computer and Information Technology (ICCIT)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICCITECHN.2017.8281778","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 2

摘要

现有的彩色图像升级算法有最近邻算法、双三次算法和双线性算法。本文首先考察了这三种算法的性能，发现双三次算法在结构相似度和执行时间方面表现最好。为了改进结果，提出了一种基于双三次反向传播的人工神经网络方法。与纯双立方型相比，人工神经网络双立方型的SSIM高6.5%，PSNR高6.9%，SNR高8.7%，MSE低30.23%。将双次神经网络的结果与SRCNN等先进的超分辨率技术进行了比较。ANN的双立方比SRCNN的SSIM高1.48%，PSNR高3.44%。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

查看原文本刊更多论文

Single image super-resolution using back-propagation neural networks

There are several existing mathematical algorithms for colour image upscaling like Nearest Neighbour, Bicubic and Bilinear. This paper firstly investigates the performances of these three and it has been found that Bicubic performs the best in terms of structural similarity and execution time. A Bicubic with backpropagation based ANN method has been proposed to improve the results. Bicubic with ANN shows 6.5% higher SSIM, 6.9% higher PSNR, 8.7% higher SNR and 30.23% lower MSE values than pure Bicubic. The results of Bicubic with ANN are also compared with state of the art super-resolution techniques like SRCNN. Bicubic with ANN produces 1.48% higher SSIM and 3.44% higher PSNR than SRCNN.

求助全文

通过发布文献求助，成功后即可免费获取论文全文。去求助

来源期刊

2017 20th International Conference of Computer and Information Technology (ICCIT)

自引率

0.00%

发文量