Image processing using cellular neural networks based on multi-valued and universal binary neurons

Multi-valued and universal binary neurons (MVN and UBN) are the neural processing elements with the complex-valued weights and high functionality. It is possible to implement an arbitrary mapping described by partially defined multiple-valued function on the single MVN. An arbitrary mapping described by partially defined or fully defined Boolean function, which can be non-threshold, may be implemented on the single UBN. The quickly converging learning algorithms exist for both types of neurons. Such features of the MVN and UBN may be used for solving the different problems. One of the most successful applications of the MVN and UBN is their usage as basic neurons in the Cellular Neural Networks (CNN). It opens the new effective opportunities in nonlinear image filtering and its applications to noise reduction, edge detection and solving of the super resolution problem. A number of experimental results are presented to illustrate the performance of the proposed algorithms.An erratum to this article can be found at     

Damage and fracture evaluation of granular composite materials by digital image correlation method

This paper presents the applications of digital image correlation technique to the mesoscopic damage and fracture study of some granular based composite materials including steelfiber reinforced concrete, sandstone and crystal-polymer composite. The deformation fields of the composite materials resulted from stress localization were obtained by the correlation computation of the surface images with loading steps and thus the related damage prediction and fracture parameters were evaluated. The correlation searching could be performed either directly based on the gray levels of the digital images or from the wavelet transform (WT) coefficients of the transform spectrum. The latter was developed by the authors and showed higher resolution and sensitivity to the singularity detection. Because the displacement components came from the rough surfaces of the composite materials without any coats of gratings or fringes of optical interferometry, both surface profiles and the deformation fields of the composites were visualized which was helpful to compare each other to analyze the damage of those heterogeneous materials. The project supported by the National Natural Science Foundation of China (10125211 and 10072002), the Scientific Committee of Yunnan Province for the Program of Steel Fiber Reinforced Concrete, and the Institute of Chemical Materials, CAEP at Mianyang     

基于公共存储器进行数据交换的高速数字图象处理系统

本文介绍了一种基于公共存储器进行数据交换的高速数字图象处理系统．给出了公共存储器的设计，描述了处理器对公共存储器的进程同步控制．     

钢卷尺刻度在线测量系

本文介绍了一种钢卷尺刻度在线测量系统 ,它采用了高速摄像及计算机数据处理技术 ,文章阐述了系统的非接触动态测量原理 ,描述了光电信号的处理方法 .     

纳米分辨率的三维形貌检测仪

从光干涉方式，瞳窗关系和光源带宽等基本关系出发，利用空间不变系统理论解析了相移显微干涉检测系统中衍射效应的影响。为了减小横向分辨率对纵向分辨率的影响，利用相关信息提取或数字滤波的方法，获得了纳米分辨率的三维形貌。     

对象随型粗糙面近红外单目视觉测量与基准图像自愈

针对复杂应用环境中近红外单目视觉位姿测量系统偏离预置合作目标对象后,无法完成位姿测量的特殊情况,提出一种以合作目标周边随型粗糙面图像为测量对象的单目视觉测量方法,以及图像受损后的动态自愈方法。通过将实时获取的随型粗糙面图像特征与预存基准图像特征进行匹配计算,完成特殊情况下的应急测量。同时,为减少随型粗糙面图像污染或受损后对位姿测量精度的影响,实时计算污染或受损程度并动态自愈基准图像特征。实验结果表明,以随型粗糙面为对象的位姿测量精度稍低于合作目标对象,但能够满足特殊情况下的应急使用需求,提高了测量系统的鲁棒性;当随型粗糙面图像污染或受损达到70%时,采用自愈处理与未做自愈处理相比,方位角测量误差减少72%以上,验证了基准图像自愈方法的有效性。     

An Image Compression Encryption Algorithm Based on Chaos and ZUC Stream Cipher

In order to improve the transmission efficiency and security of image encryption, we combined a ZUC stream cipher and chaotic compressed sensing to perform image encryption. The parallel compressed sensing method is adopted to ensure the encryption and decryption efficiency. The ZUC stream cipher is used to sample the one-dimensional chaotic map to reduce the correlation between elements and improve the randomness of the chaotic sequence. The compressed sensing measurement matrix is constructed by using the sampled chaotic sequence to improve the image restoration effect. In order to reduce the block effect after the parallel compressed sensing operation, we also propose a method of a random block of images. Simulation analysis shows that the algorithm demonstrated better encryption and compression performance.     

数字图像三点中值去隔行技术研究

对传统三点中值去隔行算法进行改进,采用同步更新迭代的方法,提高了图像的画面质量.实验证明,该算法不仅降低了时间复杂度,保护图像边缘,而且明显改善了噪声滤波效果.     

基于知识蒸馏与改进ViT网络的花卉图像细粒度分类

由于自然条件下拍摄的花卉图像背景复杂,而且其存在类内差异性大和类间相似性高的问题,现有主流方法仅依靠卷积模块提取花卉的局部特征难以实现准确的细粒度分类。针对上述问题,本文提出了1种高精度、轻量化的花卉分类方法(ConvTrans-ResMLP),通过结合Transformer模块和残差MLP(multi-layer perceptron)模块实现对花卉图像的全局特征提取,并在Transformer模块中加入卷积计算使得模型仍保留提取局部特征的能力;同时,为了进一步将花卉分类模型部署到边缘设备中,本研究基于知识蒸馏技术实现对模型的压缩与优化。实验结果表明,本文所提出的方法在Oxford 17、Oxford 102和自制的Flowers 32数据集上的准确率分别达98.62%、97.61%和98.40%;知识蒸馏后本文的轻量化模型的大小约为原来的1/18,而准确率仅下降2%左右。因此,本研究能较好地提升边缘设备下花卉细粒度分类的效率,对促进花卉培育的自动化发展具有切实意义。     

基于RGBD数据与深度权重的压缩光场显示

使用多层空间光调制器的压缩光场(compressive light field,CLF) 显示技术具有空间带宽利用率高、图像分辨率好等优势,是一种很有前景的光场显示技术。传统的方法将光场分解视为超定问题并使用优化算法求解。随着重建光场的分辨率、深度、视角等性能参数的提升,优化算法的计算效率低、内存消耗大的缺陷被放大,难以实现快速计算。为此,本文提出了一种新型CLF分解算法,该算法将原始光场分解为物点进行存储,仅占用少量内存;利用深度权重和加权平均代替了优化算法,大幅提高了运算效率。在同等条件下,所提出的算法占用内存仅是传统方法的38.8%,计算时间缩短了93.7%,图像质量提高约1 dB。最后通过仿真和实验两种方式对比了该方法和传统方法的显示效果,验证了算法的有效性。