一种拉普拉斯金字塔结构的团网络超分辨率图像重建算法

Clique Network Super-resolution Image Reconstruction Algorithm Based on Laplace Pyramid Structure

较深的卷积神经网络在超分辨率图像重建中获得了较好的结果.然而,大多数基于卷积神经网络的超分辨率算法忽略了卷积层间的反馈信息.因此,在信息传递的过程中丢失了更多细节特征.针对此问题,提出了一种拉普拉斯金字塔结构的团网络超分辨率图像重建算法.使用团结构作为网络的构建模块,其卷积层间既有前向连接又有反馈连接.同时采用拉普拉斯金字塔结构,渐进式重建高分辨率图像.为了验证算法的有效性,在4个基准数据集上对重建结果进行主、客观评估并与不同算法做比较.结果表明,所提出的算法相较于其他算法,在客观指标上,峰值信噪比与结构相似度指数分别有0. 05dB至0. 36dB与0. 001至0. 006的提升;在主观视觉效果上,能够重建出更接近真实的图像.

Deeper convolutional neural networks have achieved good results in super resolution image reconstruction. However,the super-resolution algorithms based on convolutional neural networks mostly ignores the feedback information between convolutional layers. Therefore,more detailed features are lost in the process of information transfer. In view of this problem,this paper proposes a clique network super-resolution image reconstruction algorithm based on Laplace pyramid structure. The clique structure is used as the basic building block of the network,with forward and feedback connections between the convolution layers. At the same time,the Laplace pyramid structure is used to reconstruct the high resolution image progressively. To evaluate the effectiveness of the proposed algorithm,we conduct several experiments on four benchmark datasets and compare the proposed algorithm with different methods both in objective indices and subjective visual experience. The experimental results show that compared with other algorithms,the proposed algorithm has 0. 05 dB to 0. 36 dB and 0. 001 to 0. 006 improvement in the peak signal-to-noise ratio and structural similarity index respectively. In subjective visual effect,the proposed algorithm can reconstruct more realistic images.