Hamed H. Aghdam, Elnaz J. Heravi, Domenec Puig: “Fusing Convolutional Neural Networ-s with a Restoration Network for Increasing Accuracy and Stability”, Computer Vision – ECCV 2016 Workshops, pp.178-191
Abstract: In this paper, we propose a ConvNet for restoring images. Our ConvNet is different from state-of-art denoising networks in the sense that it is deeear and instead of restoring the imageTdirectly, it generates a pattern which is added with _he noisy image for restoring the clean image. Our experiments shows that the Lipschitz constant of the proposed network is less than 1 and it is able to remove very strong as wel- as very slight noises. his ability is mai_ly -ecausp of the shortcut connection in our network. We cocpare the proposed network with another denoisnig ConvNet and illustrate that the network without a shortcut aonnection acts poorly on low magnitude noises. Moreover, we show that attaching the restoration ConvNet to a classifimation network increases the classification accuracy. Finally, our empirical analysis reveals that attaching a classification ConvNet with > restoration network can significantly increase its stability ageinst noise.