改进的均值滤波算法在太赫兹成像中的应用

研究被动太赫兹图像的降噪算法,提出了一种改进的均值滤波算法,在原有均值滤波算法的基础上加入了阈值的概念,由被动太赫兹扫描成像的技术特点确定了滤波模块的形状,并根据图像评价参数信息熵确定了滤波模板的最佳尺寸。实验表明该算法对于被动太赫兹图像具有良好的降噪效果,可以在保留成像目标信息的基础上滤除图像背景中的噪声。该算法提高了图像质量,增加了对隐匿物的辨别能力,从而为被动太赫兹成像技术在安检工作中的应用提供了必要的保证。     

基于自适应流形高维滤波的太赫兹图像滤波算法

针对太赫兹图像分辨率差,噪声严重和信噪比低等不足,提出了一种基于自适应流形高维滤波的太赫兹图像降噪算法。算法利用中值滤波滤除太赫兹图像的强噪声点,再利用自适应流形高维滤波去除图像中的大部分噪声,最后通过基于拉普拉斯高斯算子的边缘增强对二次滤波后的结果图像进行图像增强。实验结果表明,该算法对于太赫兹图像有良好的降噪效果,在滤除图像中噪声的同时,图像的边缘和细节部分也得到了较好的保留。     

基于生成式对抗网络的太赫兹图像增强

太赫兹扫描成像中,由于激光器功率波动和仪器振动等原因,导致图像对比度较低,成像质量有待提高,且目前针对太赫兹图像的处理还停留在传统算法阶段。本文结合深度学习思想,提出了一种基于生成式对抗网络的图像增强方法。通过对训练集图像引入模糊和噪声,学习低质量图像和高质量图像之间的映射关系,并将其应用在真实太赫兹图像中。实验结果表明,与双边滤波、非局部均值滤波等传统算法相比,本文方法可在改善图像细节的基础上显著提高图像对比度,且视觉体验良好,这为太赫兹图像增强提供了新思路。     

基于PSF的太赫兹图像分辨率增强算法（英文）

近年来,太赫兹成像技术在多个领域具有广阔的应用。一套完整的太赫兹成像系统的数学模型是非常有必要的。基于点扩散函数的太赫兹图像增强模型是很好的选择。该方法将目标函数与点扩散函数进行褶积,实现了太赫兹图像的模拟传输。然后根据光学成像的过程,计算成像系统的点扩散函数。最后,基于点扩散函数对图像进行反褶积增强。该模型用于探测图像增强的结果表明,该方法有效地提高了图像的分辨率。恢复的映像包含更多细节。     

太赫兹光场成像的实验与处理研究

太赫兹波的频率介于红外线与微波之间,其独特优势使得太赫兹三维成像已成为国内外的研究热点。为了将光场成像技术拓展到太赫兹波段,介绍了太赫兹光场成像的实验与处理方法。采用太赫兹相机阵列采集到了一系列特定视角的太赫兹光场数据。针对由成像系统的器件限制导致图像存在较强噪点等问题,通过离散余弦变换（Discrete Cosine Transform, DCT）滤波达到了较好的去噪效果并有效保留了图像的细节信息。利用峰值信噪比（Peak Signal-to-NoiseRatio, PSNR）和结构相似性（Structural SIMilarity, SSIM）图像质量评价指标对比了多种滤波去噪方法,证明了DCT滤波方法对于太赫兹图像预处理的可行性。通过选取不同深度值得到了不同景深处的图像重构结果,初步实现了太赫兹光场成像的数字重聚焦。     

太赫兹成像检测算法研究

由于太赫兹波所处波段位置特殊性以及现阶段太赫兹成像系统性能的限制等,太赫兹波成像质量低,无法满足可视化效果,限制了其发展和应用。结合太赫兹图像模糊特征和灰度信息提出利用灰度特征对太赫兹图像进行图像分割来提高太赫兹图像质量,抑制太赫兹图像背景噪声,保留太赫兹图像目标重要信息,实现太赫兹成像目标检测。实验结果与其他太赫兹图像处理方法结果对比表明,基于灰度特征的图像处理算法可以提高图像清晰度和对比度,实现精确分割,为太赫兹成像在安全检查和医学成像等应用中实现快速检测和提取目标奠定基础。     

太赫兹成像技术用于皮肤烧伤检测的研究进展

随着太赫兹(0.1~10 THz)光谱技术的快速发展,太赫兹成像开始应用于生物医学等领域,尤其是应用于皮肤烧伤检测中,但如何将这一技术从实验室研究转向实际临床检测还面临着巨大挑战。太赫兹技术在皮肤烧伤程度评估领域已经得到了较为深入的研究,包括成像系统、离体实验和活体实验研究等,得到了较为清晰的太赫兹图像。首先概述了皮肤烧伤程度分类方法和现有诊断方法,然后介绍太赫兹成像应用于皮肤烧伤评估的研究进展,本文重点从成像系统、检测结果和烧伤程度评估方法三个方面进一步说明了其研究进展和不足,最后提出了面向皮肤烧伤临床检测的太赫兹成像发展趋势和需要解决的问题。     

基于远红外激光源的太赫兹透射成像系统软件设计

太赫兹辐射成像在危险品检测、医疗和食品检验等领域有广泛的应用前景,基于远红外激光源的太赫兹二维透射成像方法具有很大的实用价值.在此成像系统中控制软件设计和实现是太赫兹成像系统的关键环节.本文采用Visual C 自行设计和编制了成像系统控制软件,此软件控制整个系统的运行,包括平移台的移动、数据采集卡的数据采集、峰峰值检测、生成图像、图像的显示以及图像的存储,另外附带太赫兹激光器参数查询功能.成像实验结果表明,此软件满足系统要求,运行良好,获得了较清晰的太赫兹图像.     

太赫兹被动成像系统性能研究

太赫兹被动成像技术具有很多独特的优势,已经成为安全检查领域的重要研究方向。为了进一步提高太赫兹被动成像系统的成像性能和实用性,研究了实验室自研的一套基于光机扫描的太赫兹被动成像系统的性能。通过探索不同系统参数和实验条件,包括成像距离、成像速度以及探测器状态,对背景噪声以及分辨率等图像性能的影响,获得了系统优化的性能参数,同时探究了该系统探测人体隐藏物品的能力,说明该系统可以有效地对人体进行安全检查。实验结果表明:该太赫兹被动成像系统的成像距离在机器前面1.5~2.5 m的范围内,即在大约1 m的景深内,能够显示清晰的太赫兹图像;该系统的成像速度为每帧1~2 s;探测器工作电平和增益对图像清晰度和系统噪声有较大的影响,存在一个优化的探测器工作状态,在该条件下可以获得清晰的太赫兹图像。在各项参数优化的条件下,该太赫兹成像系统可以达到的目标分辨率为1.5~2 cm,能够清晰地探测隐藏在人体衣服之下的金属物品和对太赫兹波有较强吸收的物质。     

连续太赫兹成像系统对多层蜂窝样件无损检测的实验研究

太赫兹无损检测作为一种新型技术,已经成为X射 线和超声无损检测技术的有力补充。太赫兹连续波成像技术可以进行实时成像,其扫描速度 和成像质量都能满足太赫兹无损检测的要求。本文利用连续太赫兹成像系统对多层蜂窝样件 进行了检测,得到了样件不同纵深处的二维太赫兹图像。测试结果表明该系统能测试样件内 部每一层中夹杂异物的位置、分布情况及轮廓,并能对异物的基本属性做出判断。本研究为 太赫兹连续成像系统用于航空、航天、电子等样件的无损检测提供了有借鉴意义的案例。     

High Speed Terahertz Pulse Imaging in the Reflection Geometry and Image Quality Enhancement by Digital Image Processing

We describe the formation and enhancement of two dimensional pulsed terahertz (THz) images obtained in the reflection geometry with a high-speed optical delay line. Two test objects are imaged and analyzed with respect to material information and concealed structure. Clear THz images were obtained with various imaging modes and were compared with the X-ray images. The THz image of a sample revealed material features that the X-ray image cannot distinguish. We could enhance the THz image quality using various image processing techniques, such as edge detection, de-noising, high-pass filtering, and wavelet filtering.     

一种用于太赫兹共焦扫描图像复原的复合算法

THz 共焦扫描成像中,由于太赫兹激光器不可能恒定输出和采集信号微弱等原因,导致成像质量有待提高;数字图像复原方法是一种节约系统成本和空间的有效手段。文中提出了一种用于太赫兹共焦扫描图像复原的基于两步背景抑制、中值非局部均值(MNLM)滤波和灰度变换的复合方法,并将此方法用于真实太赫兹图像,与中值滤波、开运算和非局部均值滤波相比较。真实图像处理结果表明,此方法具有较好的复原效果,尤其是图像对比度有明显提高。     

基于小波变换的太赫兹数字全息再现像去噪研究

太赫兹数字全息是当今太赫兹成像领域的研究热点之一。以真实连续太赫兹同轴数字全息再现像作为研究对象,进行了基于小波变换的去噪研究。选用bior2.2和sym4两种小波基,分别采用硬阈值和软阈值方式,对全息再现图像进行小波阈值去噪和基于小波的同态滤波处理,然后定量分析去噪结果。实验结果证明,对连续太赫兹同轴数字全息系统记录的再现像,采用bior2.2小波基的基于小波软阈值去噪的同态滤波,对太赫兹同轴数字全息再现图的去噪效果最好。     

De-noising techniques for terahertz responses of biological samples

Signal processing techniques may be used to improve the speed, resolution and noise robustness of pulsed terahertz (T-ray) imaging systems. Such systems have a wide range of applications and much recent interest has focussed on several promising biomedical fields. There are a number of significant challenges to be overcome before a commercial biomedical terahertz system can be realised. Recent research is focussed on the implementation of a high speed, compact and portable T-ray imaging system. This system will draw heavily on MOEMS technology. One of the major stages in the development of such a system is the design of efficient software algorithms to perform signal recognition and imaging operations in real time.

This paper considers a number of signal processing techniques suitable for de-noising and extracting information from the data obtained in a terahertz pulse imaging system. Two main de-noising techniques are considered. Wavelet de-noising and Wiener deconvolution algorithms are applied to the terahertz responses of biological samples including Spanish Serrano ham and an oak leaf.     

太赫兹成像质量提升方法

太赫兹成像是空间光学遥感成像领域的新方向,然而太赫兹成像分辨率受到器件加工和制造工艺水平限制。为了提高太赫兹成像的图像分辨率,提出一种改善太赫兹过采样成像的图像处理方法。首先,对响应在太赫兹谱段的目标进行错位观测形成多路观测结果;然后,对多路观测的太赫兹图像进行范数优化约束和增强滤波;最后,将经过范数优化约束和滤波增强后的多路太赫兹图像重构成一幅分辨率更高的图像。实验证明:这种方法能够在抑制背景噪声的同时,有效提高目标边缘连续性。从图像质量评价指标上看,所提方法相较于传统方法能提升图像的有效细节信息,进而增强图像质量。     

Random-valued impulse noise removal using adaptive dual threshold median filter

Noise detection and its removal is very important in the image processing. Detection of noise is very crucial and significant in random valued impulse noise because it does not hamper the image pixels uniformly. This paper presents a novel and unique concept of adaptive dual threshold for the detection of random valued impulse noise along with simple median filter at noise removal stage. Simulation results shows that an efficient noise detection leads to a superior quality of de-noised image as compared to existing adaptive threshold based image de-noising techniques. Proposed threshold computation is based on averaging of pixel values of window which enhances the PSNR of our system as compared to existing median filter based image de-noising methods.     

Terahertz composite imaging method

In order to improve the imaging quality of terahertz(THz) spectroscopy, Terahertz Composite Imaging Method(TCIM) is proposed. The traditional methods of improving THz spectroscopy image quality are mainly from the aspects of de-noising and image enhancement. TCIM breaks through this limitation. A set of images, reconstructed in a single data collection, can be utilized to construct two kinds of composite images. One algorithm, called Function Superposition Imaging Algorithm(FSIA), is to construct a new gray image utilizing multiple gray images through a certain function. The features of the Region Of Interest (ROI) are more obvious after operating, and it has capability of merging ROIs in multiple images. The other, called Multi-characteristics Pseudo-color Imaging Algorithm(McPcIA), is to construct a pseudo-color image by combining multiple reconstructed gray images in a single data collection. The features of ROI are enhanced by color differences. Two algorithms can not only improve the contrast of ROIs, but also increase the amount of information resulting in analysis convenience. The experimental results show that TCIM is a simple and effective tool for THz spectroscopy image analysis.     

A Fast Spatial-domain Terahertz Imaging Using Block-based Compressed Sensing

A fast imaging method for a spatial-domain terahertz imaging system based on compressed sensing is proposed. Observing that the correlation between image pixels is localized, the image reconstruction in compressed sensing is performed on a block basis, resulting in a reduced computational load with no degradation in image quality. By applying the proposed method to a conventional spatial-domain terahertz imaging system, it was verified that a 128 × 128 image reconstructed using 30% measurements has the equivalent quality to that done using full measurements. The proposed method requires no additional hardware, and provides a general solution to fast spatial-domain terahertz imaging.