多局部残差连接注意网络的图像去模糊

针对现有的基于卷积神经网络的图像去模糊算法存在图像纹理细节恢复不清晰的问题，提出了一种基于多局部残差连接注意网络的图像去模糊算法。首先，采用一个卷积层进行浅层特征提取；其次，设计了一种新的基于残差连接和并行注意机制的多局部残差连接注意模块，用于消除图像模糊并提取上下文信息；再次，采用一个基于扩张卷积的成对连接模块进行细节恢复；最后，利用一个卷积层重建清晰图像。实验结果表明：在GoPro数据集上的PSNR (peak signal to noise ratio)和SSIM (structure similarity)分别为31.83 dB、0.927 5，在定性和定量两方面都表明所提方法能够有效地恢复模糊图像的纹理细节，网络性能优于对比方法。

Image deblurring based on multiple local residual connection attention network

Aiming at the problem that existing image deblurring algorithms based on convolutional neural network are not clear in the restoration of image texture details, an image deblurring algorithm based on multiple local residual connection attention network was proposed. Firstly, a convolutional layer was used to extract the shallow features. Secondly, a new multiple local residual connection attention module based on residual connection and parallel attentional mechanism was designed to eliminate the image blur and extract the context information. Moreover, a pairwise connection module based on dilated convolution was adopted to restore details. Finally, a convolutional layer was used to reconstruct the clear images. The experimental results show that the peak signal to noise ratio (PSNR) and structure similarity (SSIM) on GoPro data set are 31.83 dB and 0.927 5, respectively. Both qualitative and quantitative results show that the proposed method can effectively restore the texture details of blurred images, and the network performance is better than that of the comparison method.