Pixel level image fusion scheme based on accumulated gradient and PCA transform

A new image fusion scheme based on local accumulated gradient and PCA transform is presented. The image fusion process begins by taking the Dual tree complex wavelet transform of the source images. Then, the fused local detailed coefficients are chosen according to local accumulated gradient rule, and the fused approximate coefficients are acquired by selection or averaging rule based on PCA. Finally, the fused image can be obtained by taking the inverse dual tree complex wavelet transform of the fused composite coefficients. The experimental results show that the proposed method is an efficient approach to image fusion.     

医院信息系统(HIS)系统备份与容错设计

根据医院的业务量,设计了该院的信息系统的备份与容错设计方案,采用系统冷备份与双机热备份的容错设计。运行情况表明,能充分保证医院的关键数据的安全、完整。     

基于双解码路径DD-UNet的脑肿瘤图像分割算法

针对医学图像中病灶区域尺度不一、边界模糊和周围组织强度不均匀所导致的分割精度降低问题,提出了一种基于双解码器的脑肿瘤图像分割模型。为了增强特征的表征力,提出了高阶微分残差模块并使用不同空洞率的扩张卷积用于提取特征编码,提高了网络模型的分割性能;引入上下文语义信息感知模块(multi scale dilation, MSD),从不同的目标尺度中提取更多的精细信息,提高了对结构细节信息的捕获能力,同时减少了编解码器之间的特征差异;在空间解码路径中使用选择性聚合空间注意力模块(spatial aggregation attention module, SAAM),增加了对有效空间特征的权重比例,减少了无效的特征干扰。在脑肿瘤数据集上进行了实验验证,实验结果表明,所提算法的Dice系数、平均交并比、敏感性、特异性、准确率等指标分别为：93.35%、90.71%、91.15%、99.94%、96.75%。     

基于改进3D UNet的前列腺MR图像分割

针对前列腺磁共振 (magnetic resonance, MR)图像边缘模糊、对比度较低,灰度值分布不均衡而导致分割精度较差的问题,提出了一种结合双路径注意力(dual path attention,DPA) 和多尺度特征聚合(multi-scale feature aggregation,MFA) 模块的改进3D UNet网络模型。首先,对数据集进行重采样和裁剪处理以适应模型输入。然后,在3D UNet网络的编码器各层引入DPA 并添加残差连接,加强特征的 编码能力。同时,在网络解码器中加入MFA模块,以充分利用空间上下文信息,增强语义信息。最后,在公开数据集PROMISE12上进行验证,所提出的模型的Dice系数为89.90%,Hausdorff 距离为9.37 mm。相比较于其他模型,所提出模型的分割结果更优,且参数量和运算量更少。     

Aperture-Adjustable and Repairable Metal-Based MOFs Catalysis Membrane for Water Purification

Precise adjustment of the pore size, damage repair, and efficient cleaning is all challenges for the wider application of inorganic membranes. This study reports a simple strategy of combining dry-wet spinning and electrosynthesis to fabricate stainless-steel metal–organic framework composite membranes characterized by customizable pore sizes, targeted reparability, and high catalytic activity for membrane cleaning. The membrane pore size can be precisely customized in the range of 14–212 nm at nanoscale, and damaged membranes can be repaired by targeted treatment in 120 s. In addition, advanced oxidation processes can be used to quickly clean the membrane and achieve 98% flux recovery. The synergistic actions of the membrane matrix and the selective layer increase the adsorption energy of active sites to oxidant, shorten the electron transfer cycle, and enhance the overall catalytic performance. This study can provide a new direction for the development of advanced membranes for water purification and high-efficiency membrane cleaning methods.     

结合感知损失与双重对抗网络的低剂量CT图像去噪

低剂量计算机断层扫描（LDCT）成像技术在医学诊断中得到广泛应用,但其斑纹噪声和非平稳条纹伪影复杂,目前多数算法仅依靠推断条件后验概率来实现图像去噪,无法应对LDCT图像噪声复杂、数据量少、先验知识缺乏的问题。提出一种结合感知损失的双重对抗网络去噪算法,以实现LDCT图像复原。该算法包含一个去噪器和一个生成器,分别从图像去噪和噪声生成2个角度来建模干净-噪声图像对的联合分布,通过联合学习使得去噪器和生成器相互指导,从而充分学习数据中的噪声信息和清晰图像信息,且学习到的去噪器可以直接用于LDCT图像修复。考虑到通过感知损失学习语义特征差异可以使去噪结果保留更多的细节和边缘信息,提出一种掩膜自监督方法,针对CT图像域训练一个语义特征提取网络用于计算感知损失。实验结果表明,与BM3D、RED-CNN、WGAN-VGG等主流去噪算法相比,该算法可以有效抑制噪声并去除伪影,最大程度地保留边缘轮廓和纹理细节,产生更符合人眼视觉特性的去噪效果,与当下LDCT图像去噪性能较好的SACNN算法相比,所提算法的PSNR和SSIM指标分别提升1.26 dB和1.8%。     

3D Printing Thermally Stable High-Performance Polymers Based on a Dual Curing Mechanism

High-performance polymers are an important class of materials that are used in challenging conditions, such as in aerospace applications. Until now, 3D printing based on stereolithography processes can not be performed due to a lack of suitable materials. There is report on new materials and printing compositions that enable 3D printing of objects having extremely high thermal resistance, with Tg of 283 °C and excellent mechanical properties. The printing is performed by a low-cost Digital Light Processing printer, and the formulation is based on a dual-cure mechanism, photo, and thermal process. The main components are a molecule that has both epoxy and acrylate groups, alkylated melamine that enables a high degree of crosslinking, and a soluble precursor of silica. The resulting objects are made of hybrid materials, in which the silicon is present in the polymeric backbone and partly as silica enforcement particles.     

Ultraviolet photodetectors and readout based on a-IGZO semiconductor technology

In this work, real-time ultraviolet photodetectors are realized through metal–semiconductor–metal (MSM) structures. Amorphous indium gallium zinc oxide (a-IGZO) is used as semiconductor material and gold as metal electrodes. The readout of an individual sensor is implemented by a transimpedance amplifier (TIA) consisting of an all-enhancement a-IGZO thin-film transistor (TFT) operational amplifier and a switched capacitor (SC) as feedback resistance. The photosensor and the transimpedance amplifier are both manufactured on glass substrates. The measured photosensor possesses a high responsivity R , a low response time t _{R E S} , and a good noise equivalent power value NEP .     

基于模函数与像素值差值的高质量双图像可逆信息隐藏

为解决在一些一对二的交流场景中使用信息隐藏技术来传递信息时对载密图像的视觉质量和载体图像的精确度的高要求问题. 在本文中提出了一种基于模函数和像素值差值(pixel value difference , PVD)的双图像可逆信息隐藏方案, 通过模函数和对数函数确定了PVD范围表, 从而确定在单位面积上的信息嵌入位数以及模函数的系数. 所提出的方案可以在信息嵌入位数不断增加的情况下仍然保持像素值的修改量与信息嵌入位数之比不大于0.5, 所以与目前一些基于PVD的方案相比在像素对差值越大的图像中越占有优势. 实验结果表明与现有的一些在载密图像质量方面优质的方案相比, 具有更高的PSNR和SSIM, 此外本方案在抗RS隐写分析和PDH隐写分析的静态攻击方面上具有良好的性能, 并且避免了大多数在基于像素值差值的信息隐藏方案中对溢出问题的解决方案复杂繁琐的情况.     

基于能量回馈的电力电子变压器软启动策略

电力电子变压器(PET)中存在大量初始能量为0的电容器,若直接启动PET通过常规闭环控制给这些电容充电以建立直流电压,则将导致PET内部各个变换器的电流过流,可能会导致器件的损坏。为解决此问题,提出了一种基于能量回馈的PET软启动策略。基于PET各部分变换器的电流和功率模型,该策略能够在软启动全过程中实时计算和控制峰值电流,达到了限制启动冲击电流的目的。该策略始终令双有源桥(DAB)变换器工作于峰值电流控制模式下,在保证安全的前提下最大化利用DAB变换器的功率传输能力以加快软启动速度。该策略将低压直流母线电容适度过充,并适时将其中部分能量回馈到高压侧,使得在级联H桥变换器即将进入闭环控制前其高压直流母线电压就已达到参考值,从而彻底消除交流侧的冲击电流。所提策略避免了复杂的参数设计过程,完全依靠理论模型而不是经验来指导软启动全过程的设计。仿真与实验结果验证了所提策略的正确性和有效性。