基于多通道忆阻脉冲耦合神经网络的多帧图像超分辨率重建算法  

Multi-channel Memristive Pulse Coupled Neural Network Based Multi-frame Images Super-resolution Reconstruction Algorithm

作　　者：董哲康 杜晨杰 林辉品 赖俊昇[3] 胡小方 段书凯[4] DONG Zhekang;DU Chenjie;Lin Huipin;LAI Chun sing;HU Xiaofang;DUAN Shukai(School of Electronic Information,Hangzhou Dianzi University,Hangzhou 310018,China;College of Electrical Engineering,Zhejiang University,Hangzhou 310027,China;Department of Engineering Science,University of Oxford,Oxford OX13PJ,United Kingdom;College of Artificial Intelligence,Southwest University,Chongqing 400715,China)

机构地区：[1]杭州电子科技大学电子信息学院,杭州310018 [2]浙江大学电气工程学院,杭州310027 [3]牛津大学工程科学系,英国牛津OX13PJ [4]西南大学人工智能学院,重庆400715

出　　处：《电子与信息学报》2020年第4期835-843,共9页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61571394,61601376);浙江省属高校基本科研业务费项目(GK199900299012-010)。

摘　　要：高清晰度的图像是信息获取和精确分析的前提,研究多帧图像的超分辨率重建能够有效解决因外部拍摄环境引起的图像细节丢失、边缘模糊等问题。该文基于纳米级忆阻器,设计一种多通道忆阻脉冲耦合神经网络模型(MMPCNN),能够有效模拟网络中连接系数的动态变化,解决神经网络中固有的参数估计问题。同时,将提出的网络应用于多帧图像超分辨率重建中,实现低分辨率配准图像的融合操作,并通过基于稀疏编码的单帧图像超分辨率重构算法对获得的初始高分辨率图像进行优化。最终,一系列计算机仿真及分析(主观/客观分析)验证了该文提出方案的正确性和有效性。The high-resolution image is the prerequisite of information acquisition and precise analysis.Multiframe super-resolution images reconstruction technologies are able to address many image degraded issues(caused by external shooting environment),such as detail information lost,blurred edges,and so forth.According to the nanoscale memristor,a Multi-channel Memristive Pulse Coupled Neural Network(MMPCNN)model is proposed.This model is able to simulate the adaptive-variable linking coefficient in pulse coupled neural network.Meanwhile,the proposed network is applied to the multi-frame super resolution reconstruction for fusing the registered low resolution images.Furthermore,the sparse coding based super resolution method is performed to improve the original high-resolution image.Finally,a series of computer experiments and the relevant subjective/objective analysis jointly illustrate the validity and effectiveness of the entire scheme.

关 键 词：忆阻器 脉冲耦合神经网络 多帧图像 超分辨率重建 

分 类 号：TN601[电子电信—电路与系统] TN911.73