摘要
针对相机成像时相机抖动、物体运动等导致图像产生运动模糊这一十分具有挑战性的问题,提出基于生成对抗网络的深度卷积神经网络来复原模糊图像的解决方案。该方案省略了模糊核估计的过程,采用端对端的方式直接获取复原图像;通过引入生成对抗网络思想的对抗损失和对残差网络进行改进,有效地复原了图像的细节信息。最后通过训练此深度卷积神经网络模型并在相关模糊复原基准数据集上测试,证明了该方案取得了较好的结果。
Image motion blur is a very challenging problem caused by camera shaking or object movements. In order to tackle this problem, the paper proposes a deep convolutional neural network based on generative adversarial networks method. The proposed method can restore a clear image in an end-to-end way without estimating blur kernel. By introducing adversarial loss based on generative adversarial networks and modifying the residual network structure, the proposed method can restore image details effectively. Then this paper trains this deep convolutional neural network model on public datasets. Finally, it is proved that the proposed method achieves good results according to the test on blurry image benchmark datasets.
引文
[1]程俊廷,左旺孟.快速非均匀模糊图像的盲复原模型[J].黑龙江科技大学学报,2017,27(2):196-199.
[2]张广明,高爽,尹增山,等.基于模糊图像和噪声图像的遥感图像运动模糊复原方法[J].电子设计工程,2017,25(18):82-86.
[3]Levin A,Weiss Y,Durand Y,et al.Understanding and evaluating blind deconvolution algorithms[C]//IEEE Conference on Computer Vision and Pattern Recognition,2009:1964-1971.
[4]Xu L,Ren J,Liu C,et al.Deep convolutional neural network for image deconvolution[C]//Advances in Neural Information Processing Sytems,2014:1790-1798.
[5]Schuler C,Hirsch M,Harmeling S,et al.Learning to deblur[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,38(7):1439-1451.
[6]Couzinie-Devy F,Sun J,Alahari K,et al.Learning to estimate and remove non-uniform image blur[C]//IEEE Conference on Computer Vision and Pattern Recognition,2013:1075-1082.
[7]Sun J,Cao W,Xu Z,et al.Learning a convolutional neural network for non-uniform motion blur removal[C]//IEEE Conference on Computer Vision and Pattern Recognition,2015:769-777.
[8]Nah S,Kim T,Lee K.Deep multi-scale convolutional neural network for dynamic scene deblurring[C]//IEEE Conference on Computer Vision and Pattern Recognition,2017:3883-3891.
[9]He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition,2016:770-778.
[10]Sajjadi M,Sch?lkopf B,Hirsch M.EnhanceNet:single image super-resolution through automated texture sythesis[C]//IEEE International Conference on Computer Vision,2017:4491-4500.
[11]Radford A,Metz L,Chintala S.Unsupervised representation learning with deep convolutional generative adversarial networks[C]//International Conference on Learning Representations,2016:780-789.
[12]Ramakrishnan1 S,Pachori S,Gangopadhyay A,et al.Deep generative filter for motion deblurring[C]//IEEE International Conference on Computer Vision,2017:2993-3000.
[13]Whyte O,Sivic J,Zisserman A,et al.Non-uniform deblurring for shaken images[J].International Journal of Computer Vision,2012,98(2):168-186.
[14]Gong D,Yang J,Liu L,Zhang Y et al.From motion blur to motion flow:a deep learning solution for removing heterogeneous motion blur[C]//IEEE Conference on Computer Vision and Pattern Recognition,2017:3806-3815.
[15]Lai W,Huang J,Hu Z,et al.A comparative study for single image blind deblurring[C]//IEEE Conference on Computer Vision and Pattern Recognition,2016:1701-1709.
[16]K?hler R,Hirsch M,Mohler M,et al.Recording and playback of camerashake:benchmarking blind deconvolution with a real-world database[C]//European Conference on Computer Vision,2012:27-40.