Seite - 150 - in Joint Austrian Computer Vision and Robotics Workshop 2020

up new possibilities in areas where acquiring clean training data is too timeconsumingor infeasible. There are some limitations that we leave for future research. Due to the structure of our dataset, the num- berofsamplesavailable forN2Nlearningwas limited by the available ground truth targets. Since N2N does not require manual frame editing, it is possible to increase the size of the dataset without much effort. Along with the increase of the size of the dataset, the model complexity could be increased, typically resulting in better performance. Acknowledgements The authors acknowledge grant support from the National Institutes of Health under grant 1R01EB024532-02. References [1] P. Arias and J. Morel. Video denoising via empirical bayesian estimation of space-time patches. JMIV, 60:70–93,2018. [2] H.C.Burger,C. J.Schuler, andS.Harmeling. Image denoising: Canplainneuralnetworkscompetewith BM3D? InCVPR, 2012. [3] J. Caballero, C. Ledig, A. Aitken, A. Acosta, J. Totz, Z. Wang, and W. Shi. Real-time video super- resolution with spatio-temporal networks and motion compensation.CVPR, 2017. [4] A. Chambolle. An algorithm for total variation mini- mization and applications. JMIV, 20(1):89–97, 2004. [5] M.ClausandJ.C.vanGemert. ViDeNN:Deepblind videodenoising. InCVPRWorkshops, 2019. [6] K.Dabov,A.Foi,V.Katkovnik, andK.Egiazarian. Imagedenoisingbysparse3-d transform-domaincol- laborativefiltering. IP, 16(8):2080–2095, 2007. [7] T.Ehret,A.Davy, J.Morel,G.Facciolo, andP.Arias. Model-blind video denoising via frame-to-frame training. InCVPR, 2019. [8] S. Guo, Z. Yan, K. Zhang, W. Zuo, and L. Zhang. Toward convolutional blind denoising of real pho- tographs. InCVPR, 2019. [9] K.He,X.Zhang,S.Ren, andJ.Sun. Deep residual learning for image recognition. InCVPR, 2016. [10] S. Ioffe and C. Szegedy. Batch normalization: Ac- celeratingdeepnetwork trainingby reducing internal covariate shift. In ICML, 2015. [11] V. JainandH.Seung. Natural image denoisingwith convolutionalnetworks. InNIPS, 2008. [12] D. P. Kingma and J. L. Ba. Adam: A method for stochasticoptimization. In ICLR, 2015. [13] E. Kobler, T. Klatzer, K. Hammernik, and T. Pock. Variational networks: Connecting variational meth- odsanddeep learning. InGCPR, 2017. [14] S. Laine, J. Lehtinen, and T. Aila. Improved self- superviseddeep imagedenoising. In ICLR, 2019. [15] J. Lehtinen, J. Munkberg, J. Hasselgren, S. Laine, T. Karras, M. Aittala, and T. Aila. Noise2Noise: Learning image restoration without clean data. In ICML, 2018. [16] J. Long, E. Shelhamer, and T. Darrell. Fully convolu- tionalnetworks for semantic segmentation. InCVPR, 2015. [17] M. Maggioni, G. Boracchi, A. Foi, and K. Egiazar- ian. Videodenoising,deblocking, andenhancement through separable 4-d nonlocal spatiotemporal trans- forms. IP, 21(9):3952–3966, 2012. [18] X.-J.Mao,C.Shen, andY.-B.Yang. Image restora- tion using convolutional auto-encoders with symmet- ric skipconnections.ArXiv, abs/1606.08921,2016. [19] X.-J.Mao,C.Shen, andY.-B.Yang. Image restora- tionusingverydeepconvolutional encoder-decoder networks with symmetric skip connections. InNIPS, 2016. [20] V. Nair and G. Hinton. Relus improve restricted boltzmann machines. In ICML, 2010. [21] S. Niklaus and F. Liu. Context-aware synthesis for video frame interpolation. InCVPR, 2018. [22] Y. Roh, G. Heo, and S. E. Whang. A survey on datacollection formachine learning: Abigdata - ai integrationperspective.KDE, 2019. [23] O. Ronneberger, P. Fischer, and T. Brox. U-net: Con- volutionalnetworks forbiomedical image segmenta- tion. InMICCAI, 2015. [24] L. I. Rudin, S. Osher, and E. Fatemi. Nonlinear total variationbasednoise removal algorithms.PhysicaD, 60(1):259–268,1992. [25] S.Su,M.Delbracio, J.Wang,G.Sapiro,W.Heidrich, andO.Wang. Deepvideodeblurring forhand-held cameras.CVPR, 2017. [26] D. Sun, X. Yang, M.-Y. Liu, and J. Kautz. PWC-Net: CNNsforopticalflowusingpyramid,warping, and costvolume. InCVPR, 2018. [27] C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D.Anguelov,D.Erhan,V.Vanhoucke, andA.Rabi- novich. Goingdeeperwithconvolutions. InCVPR, 2015. [28] M. Tassano, J. Delon, and T. Veit. DVDnet: A fast Network forDeepVideoDenoising. In ICIP, 2019. [29] P. Weinzaepfel, J. Revaud, Z. Harchaoui, and C. Schmid. Deepflow: Large displacement optical flowwithdeepmatching. ICCV, 2013. [30] K.Zhang, W.Zuo, Y. Chen, D.Meng, andL. Zhang. Beyond a gaussian denoiser: Residual learning of deepCNNforimagedenoising. IP,26(7):3142–3155, 2017. [31] K. Zhang, W. Zuo, and L. Zhang. Ffdnet: Toward a fast and flexible solution for CNN based image denoising. IP, 2018. 150