Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurring

摘要: Non-uniform blind deblurring for general dynamic scenes is a challenging computer vision problem as blurs arise not only from multiple object motions but also camera shake, scene depth variation. To remove these complicated motion blurs, conventional energy optimization based methods rely on simple assumptions such that blur kernel partially uniform or locally linear. Moreover, recent machine learning depend synthetic datasets generated under assumptions. This makes fail to where difficult approximate parameterize (e.g. boundaries). In this work, we propose multi-scale convolutional neural network restores sharp images in an end-to-end manner caused by various sources. Together, present loss function mimics coarse-to-fine approaches. Furthermore, new large-scale dataset provides pairs of realistic blurry image and the corresponding ground truth are obtained high-speed camera. With proposed model trained dataset, demonstrate empirically our method achieves state-of-the-art performance qualitatively, quantitatively.