{"title":"用于盲去模糊的自监督多尺度神经网络","authors":"Meina Zhang, Ying Yang, Guoxi Ni, Tingting Wu, Tieyong Zeng","doi":"10.3934/ipi.2023046","DOIUrl":null,"url":null,"abstract":"Blurry kernel estimation is a critical yet challenging task for blind deblurring. Most existing works devote to designing end-to-end networks that require a large amount of hard-to-obtain training data. In addition, these methods often ignore the intrinsic effects of blur kernel for blind deblurring. In this work, we present a unified latent image deblur and kernel estimation method based on MAP framework. By revisiting the coarse-to-fine strategy, we introduce a self-supervised multi-scale deblur network(MD-Net), where the multi-scale structure significantly reduce the kernel deviation caused by local area minimization. Specifically, our network commences with random inputs and outputs multi-scale reconstructed images and kernels. By progressively capturing the high-level configuration and low-level details from matching multi-resolution loss functions, the proposed MD-Net enable to capture multi-level image priors. Meanwhile, at each coarse level, we use Feature Extraction(FE) layers to further extract and emphasize features from reconstructed images. Compared with state-of-the-art blind deblurring methods, extensive experiments demonstrate that the proposed approach significantly improves the restoration performance in both quantitative and qualitative evaluations.","PeriodicalId":50274,"journal":{"name":"Inverse Problems and Imaging","volume":"5 1","pages":"0"},"PeriodicalIF":1.2000,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Self-supervised multi-scale neural network for blind deblurring\",\"authors\":\"Meina Zhang, Ying Yang, Guoxi Ni, Tingting Wu, Tieyong Zeng\",\"doi\":\"10.3934/ipi.2023046\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Blurry kernel estimation is a critical yet challenging task for blind deblurring. Most existing works devote to designing end-to-end networks that require a large amount of hard-to-obtain training data. In addition, these methods often ignore the intrinsic effects of blur kernel for blind deblurring. In this work, we present a unified latent image deblur and kernel estimation method based on MAP framework. By revisiting the coarse-to-fine strategy, we introduce a self-supervised multi-scale deblur network(MD-Net), where the multi-scale structure significantly reduce the kernel deviation caused by local area minimization. Specifically, our network commences with random inputs and outputs multi-scale reconstructed images and kernels. By progressively capturing the high-level configuration and low-level details from matching multi-resolution loss functions, the proposed MD-Net enable to capture multi-level image priors. Meanwhile, at each coarse level, we use Feature Extraction(FE) layers to further extract and emphasize features from reconstructed images. Compared with state-of-the-art blind deblurring methods, extensive experiments demonstrate that the proposed approach significantly improves the restoration performance in both quantitative and qualitative evaluations.\",\"PeriodicalId\":50274,\"journal\":{\"name\":\"Inverse Problems and Imaging\",\"volume\":\"5 1\",\"pages\":\"0\"},\"PeriodicalIF\":1.2000,\"publicationDate\":\"2023-01-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Inverse Problems and Imaging\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.3934/ipi.2023046\",\"RegionNum\":4,\"RegionCategory\":\"数学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"MATHEMATICS, APPLIED\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Inverse Problems and Imaging","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.3934/ipi.2023046","RegionNum":4,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"MATHEMATICS, APPLIED","Score":null,"Total":0}
Self-supervised multi-scale neural network for blind deblurring
Blurry kernel estimation is a critical yet challenging task for blind deblurring. Most existing works devote to designing end-to-end networks that require a large amount of hard-to-obtain training data. In addition, these methods often ignore the intrinsic effects of blur kernel for blind deblurring. In this work, we present a unified latent image deblur and kernel estimation method based on MAP framework. By revisiting the coarse-to-fine strategy, we introduce a self-supervised multi-scale deblur network(MD-Net), where the multi-scale structure significantly reduce the kernel deviation caused by local area minimization. Specifically, our network commences with random inputs and outputs multi-scale reconstructed images and kernels. By progressively capturing the high-level configuration and low-level details from matching multi-resolution loss functions, the proposed MD-Net enable to capture multi-level image priors. Meanwhile, at each coarse level, we use Feature Extraction(FE) layers to further extract and emphasize features from reconstructed images. Compared with state-of-the-art blind deblurring methods, extensive experiments demonstrate that the proposed approach significantly improves the restoration performance in both quantitative and qualitative evaluations.
期刊介绍:
Inverse Problems and Imaging publishes research articles of the highest quality that employ innovative mathematical and modeling techniques to study inverse and imaging problems arising in engineering and other sciences. Every published paper has a strong mathematical orientation employing methods from such areas as control theory, discrete mathematics, differential geometry, harmonic analysis, functional analysis, integral geometry, mathematical physics, numerical analysis, optimization, partial differential equations, and stochastic and statistical methods. The field of applications includes medical and other imaging, nondestructive testing, geophysical prospection and remote sensing as well as image analysis and image processing.
This journal is committed to recording important new results in its field and will maintain the highest standards of innovation and quality. To be published in this journal, a paper must be correct, novel, nontrivial and of interest to a substantial number of researchers and readers.