Restoring Blurred Image with Capsule Network
Zengguang Tian, Wenbiao Zhang, Zuo Jiang, Yi Tang · 2021 4th International Conference on Artificial Intelligence and Pattern Recognition · 2021
Image deblurring can effectively improve image quality and machine recognition accuracy. Existing methods use universal prior information in the data to deblur, and rarely consider image semantic prior information. In the field of deep learning, using more prior information can achieve better results. Inspired by the powerful object semantic information representation ability of the capsule network, we propose a capsule neural network based on semantic information deblurring. The blurred image and the clear image have different semantic distributions in the capsule feature space. A mapping network is used to map the blurred image capsule feature distribution to the clear image capsule feature distribution, and then input the clear image training decoder to obtain the corresponding clear image. We demonstrate that our proposed method can achieve good visual effects on the MNIST and Fashion-MNIST data sets, the PSNR and SSIM evaluation indicators are better than the classic deblurring methods.