基于稀疏表示的对抗网络图像聚焦形貌修复  

Confrontation Network Image FocusShape Restoration Based on Sparse Representation

作　　者：陆焱[1] 胡玉荣 郭竞[3] LU Yan;HU Yu-rong;GUO Jing(School of Computer Engineering,Jingchu Institute of Technology,Jingmen Hubei448000,China;Department of Science and Technology,Jingchu Institute of Technology,JingmenHubei 448000,China;School of Information Science and Technology,Northwestern University,Xi'an Shanxi 710127,China)

机构地区：[1]荆楚理工学院计算机工程学院,湖北荆门448000 [2]荆楚理工学院科技处,湖北荆门448000 [3]西北大学信息科学与技术学院,西安长安710127

出　　处：《计算机仿真》2021年第3期126-130,258,共6页Computer Simulation

基　　金：2020年度湖北省教育厅科学研究计划指导性项目(B2020192);2020年荆楚理工学院科研项目(YB202007);2020年荆楚理工学院国家级培育项目(PY202003)。

摘　　要：传统方法在进行对抗网络图像聚焦形貌修复时,图像光影边缘修补效果不佳,图像去噪效果差,提出一种稀疏表示的对抗网络图像聚焦形貌修复方法。通过摄像机采集到基本的目标图像,利用计算机图像扫描程序对目标图像分别进行整体的与局部的图像扫描,采集需要修复的图像内部信息。使用图像融合法融合图像数据,提取特征点。通过预估字典矩阵确定字典类型,运用贪婪算法中的追踪匹配法,实现图像数据匹配与缺陷填补。使用平滑处理的方式改善线条与光影边缘部分,得到清晰的图像。实验结果表明,设计的图像聚焦形貌修复方法能够提高目标图像完整融合,提高图像修复清晰度,有效解决噪声问题。Generally,the traditional methods have some shortcomings,such as the poor effect of image edge repair and image denoising.Therefore,this paper proposes a sparse representation method for the restoration of focused topography of anti-network images.First of all,computer image scanning program was used to scan the whole and local image of the target image respectively,and collect the internal information of the image to be repaired.Secondly,Image fusion method was adopted to fuse image data and extract feature points.Then,based on the estimated dictionary matrix,the dictionary type was determined.The tracking matching method of greedy algorithm was utilized to achieve image data matching and defect filling.Finally,the smooth processing method was used to improve the line and light shadow edge part to obtain a clear image.The simulation results show that the method in this word effectively improves the fusion of the target image and the clarity of the image restoration,and eliminates the noise interference.

关 键 词：稀疏表示 对抗网络 网络图像 图像聚焦 形貌修复 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]