基于视觉传达的多帧图像高分辨率重建仿真  

High Resolution Reconstruction Simulation of Multi-frame Image Based on Visual Communication

作　　者：李博[1] LI Bo(China Medical University Shenyang,Shenyang Liaoning 110000,China)

机构地区：[1]中国医科大学,辽宁沈阳110000

出　　处：《计算机仿真》2021年第3期113-116,121,共5页Computer Simulation

摘　　要：针对传统的高分辨图像重建方法,重建之后的图像细节不够丰富清晰,边缘模糊的问题,提出了一种基于视觉传达的多帧图像高分辨率重建方法。采用深度学习方法提取高分辨率图像的深层次特征,在稀疏字典超分辨率框架下联合训练特征字典,将提取出来的特征视为ScSR模型中的特征样本,代入PCANet的特征字典中,以此为基础,基于稀疏正则模型对高分辨率图像进行重建,在反向投影全局优化模型基础上引入非局部近似性先验约束对重建图像进行优化,完成多帧图像高分辨率重建优化。实验结果表明,所提方法与其它传统方法相比,图像重建效果更好,图像边缘更加清晰。In traditional high-resolution image reconstruction methods,the image detail after reconstruction is not clear,and the edge is blurred.Therefore,this article puts forward a high-resolution reconstruction method for multi-frame image based on visual communication.At first,the depth learning method was used to extract the deep features from high-resolution image.Then,the characteristic dictionary was trained under the super-resolution framework of sparse dictionary.Moreover,the extracted characteristics were regarded as the feature sample in ScSR model,and then the feature sample was substituted into PCANet characteristic dictionary.On this basis,the high resolution image was reconstructed by sparse regularization model.On the basis of the backward projection global optimization model,the non-local approximation prior constraint was introduced to optimize the reconstruction image.Finally,the high-resolution reconstruction optimization of multi-frame image was completed.Simulation results show that the reconstruction effect of proposed method is better than that of traditional methods,so that the image edge is clearer.

关 键 词：视觉传达 多帧图像 高分辨率 重建 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]