Image super-resolution reconstruction network combining asymmetric convolution and feature distillation
In order to further improve the image reconstruction effect of single image super-reso-lution(SISR)lightweight network,based on lightweight network RFDN,an image super-resolu-tion reconstruction network combining asymmetric convolution and feature distillation was pro-posed.Firstly,asymmetric convolution was used to construct a feature extraction module,the asymmetric convolution of two different convolution kernels in parallel in the residual block en-hances the feature extraction capability of the network.Secondly,the balanced attention mecha-nism(BAM)and asymmetric convolution were used to improve the feature distillation module for the acquisition of high frequency information.Finally,BAM was added to the reconstruction module to further improve the final reconstruction performance of the network.The experimental results show that compared with advanced lightweight networks such as RLFN and SMSR,the proposed ACDN can reconstruct high-quality images with richer texture details on five standard data sets,improve the peak signal-to-noise ratio and structural similarity index of reconstructed images,and achieve a better balance between the number of parameters and the performance of the network model.
国家科技期刊平台
NSTL
万方数据
维普
