当前位置:文档之家› 提高超声成像的分辨率

提高超声成像的分辨率

提高超声成像的分辨率
提高超声成像的分辨率

552IEEE TRANSACTIONS ON IMAGE PROCESSING,VOL.18,NO.3,MARCH2009 Low Bit-Rate Image Compression via Adaptive Down-Sampling and Constrained Least

Squares Upconversion

Xiaolin Wu,Senior Member,IEEE,Xiangjun Zhang,and Xiaohan Wang

Abstract—Recently,many researchers started to challenge a long-standing practice of digital photography:oversampling followed by compression and pursuing more intelligent sparse sampling techniques.In this paper,we propose a practical ap-proach of uniform down sampling in image space and yet making the sampling adaptive by spatially varying,directional low-pass pre?ltering.The resulting down-sampled pre?ltered image re-mains a conventional square sample grid,and,thus,it can be compressed and transmitted without any change to current image coding standards and systems.The decoder?rst decompresses the low-resolution image and then upconverts it to the original resolu-tion in a constrained least squares restoration process,using a2-D piecewise autoregressive model and the knowledge of directional low-pass pre?ltering.The proposed compression approach of collaborative adaptive down-sampling and upconversion(CADU) outperforms JPEG2000in PSNR measure at low to medium bit rates and achieves superior visual quality,as well.The superior low bit-rate performance of the CADU approach seems to suggest that oversampling not only wastes hardware resources and energy, and it could be counterproductive to image quality given a tight bit budget.

Index Terms—Autoregressive modeling,compression standards, image restoration,image upconversion,low bit-rate image com-pression,sampling,subjective image quality.

I.I NTRODUCTION

O VER the past half century,the prevailing engineering practice of image/video compression,as exempli?ed by all existing image and video compression standards,is to start with a dense2-D sample grid of https://www.doczj.com/doc/882066110.html,pression is done by transforming the spatial image signal into a space(e.g.,spaces of Fourier or wavelet bases)in which the image has a sparse rep-resentation and by entropy coding of transform coef?cients.Re-cently,researchers in the emerging?eld of compressive sensing challenged the wisdom of what they called“oversampling fol-lowed massive dumping”approach.They showed,quite sur-prisingly,it is possible,at least theoretically,to obtain compact signal representation by a greatly reduced number of random samples[1].

Manuscript received March12,2008;revised November05,2008.Current version published February11,2009.The associate editor coordinating the re-view of this manuscript and approving it for publication was Prof.Bruno Car-pentieri.

The authors are with the Department of Electrical and Computer Engineering, McMaster University,Hamilton,ON L8S4K1Canada(e-mail:xwu@ece.mc-master.ca;zhangxj@grads.ece.mcmaster.ca;wangx28@mcmaster.ca).

Color versions of one or more of the?gures in this paper are available online at https://www.doczj.com/doc/882066110.html,.

Digital Object Identi?er10.1109/TIP.2008.2010638

In this paper,we investigate the problem of compact image representation in an approach of sparse sampling in the spa-tial domain.The fact that most natural images have an expo-nentially decaying power spectrum suggests the possibility of interpolation-based compact representation of images.A typ-ical scene contains predominantly smooth regions that can be satisfactorily interpolated from a sparsely sampled low-resolu-tion image.The dif?culty is with the reconstruction of high fre-quency contents.Of particular importance is faithful reconstruc-tion of edges without large phase errors,which is detrimental to perceptual quality of a decoded image.

To answer the above challenge,we develop a new image com-pression methodology of collaborative adaptive down-sampling and upconversion(CADU).The CADU approach puts an em-phasis on edge reconstruction for the perceptual importance of edges and also to exploit the anisotropy of edge spectrum.In the CADU encoder design,we choose not to perform uneven irregular down sampling of an input image according to local spatial or frequency characteristics.Instead,we stick to con-ventional square pixel grid by uniform spatial down sampling of the image.Consequently,the resulting low-resolution image can be readily compressed by any of existing image compres-sion techniques,such as DCT-based JPEG and JPEG2000.Yet the simple uniform down-sampling scheme is made adaptive by a directional low-pass pre?ltering step prior to down-sampling. The other purpose of this preprocessing is to induce a mecha-nism of collaboration between the spatial uniform down-sam-pling process at the encoder and optimal upconversion process at the decoder.

The CADU decoder?rst decompresses the low-resolution image and then upconverts it to the original resolution in con-strained least squares restoration process,using a2-D piece-wise autoregressive model(PAR)and by reversing the direc-tional low-pass pre?ltering operation of the encoder.Two-di-mensional autoregressive modeling was a known effective tech-nique of predictive image coding[2],[3].For the CADU de-coder,the PAR model plays a role of adaptive noncausal pre-dictor.What is novel and unique of the CADU approach is that the predictor is only used at the decoder side,and the noncausal predictive decoding is performed in collaboration with the pre-?ltering of the encoder.

Fig.1presents a block diagram of the proposed CADU image compression system,summarizing the above ideas and depicting the collaboration between the encoder and decoder. The CADU image compression technique,although oper-ating on down-sampled images,obtains some of the best PSNR

1057-7149/$25.00?2009IEEE

WU et al.:LOW BIT-RATE IMAGE COMPRESSION VIA ADAPTIVE DOWN-SAMPLING553 results and visual quality at low to medium bit rates.CADU

outperforms the JPEG2000standard,even though the latter is

fed images of higher resolution and is widely regarded as an

excellent low bit-rate image codec[4].

Since the down-sampled image has the conventional form of

square pixel grid and can be fed directly to any existing image

codec,standard or proprietary,the CADU upconversion process

is entirely up to the decoder.Thus,as shown in Fig.1,the pro-

posed CADU image coding approach can work in tandem with

any third party image/video compression techniques.This?exi-

bility makes standard compliance a nonissue for the new CADU

method.We envision that CADU becomes a useful enhancer

of any existing image compression standard for improved low

bit-rate performance.

An important application of CADU is visual communica-

tion in wireless networks,where the bandwidth is at a premium

and end devices have diverse display capabilities,ranging from

small screens of cell phones to regular screens of laptops.As

illustrated in Fig.1,the same standard-compliant code

stream

feeds displays of different resolutions.The only difference is that high-resolution displays invoke CADU upconversion while low-resolution displays do not.This design happens to be com-patible with network hardware levels since high-resolution dis-plays are typically associated with more powerful computers.In such a heterogenous wireless environment scalable or layered image compression methods(e.g.,JPEG2000)are inef?cient, because the re?nement portion of the scalable code stream still consumes bandwidth and yet generates no bene?ts to low-reso-lution devices.

Because the down-sampled image is only a small fraction of the original size,CADU greatly reduces the encoder com-plexity,regardless what third-party codec is used in conjunc-tion.This property allows the system to shift the computation burden from the encoder to decoder,making CADU a viable asymmetric compression solution when the encoder is resource deprived.Furthermore,the superior low bit-rate performance of the CADU approach seems to suggest that a camera of unnec-essarily high resolution can ironically produce inferior images than a lower resolution camera,if given a tight bit budget.This rather counter-intuitive observation should be heeded when one designs visual communication devices/systems with severe con-straints of energy and bandwidth.

Closely related to our work is a series of recent papers on improving the performance of DCT-based JPEG image compression standard at low bit rates by adaptive downsam-pling at encoder and interpolation at decoder[5]–[8].In[5], Bruckstein et al.explained analytically why it is advantageous to downsample an image prior to JPEG compression and upsample the JPEG-decoded image.The authors developed a model for expected distortion of DCT-based JPEG codec and gave an expression to determine the rate-distortion optimal downsampling factor.Following up on[5],Tsaig et al.pro-posed an image-dependent algorithm to?nd optimal?lters for decimation and interpolation[6].Lin and Dong studied the so-called critical bit rate,below which it pays to downsample an image prior to DCT-based JPEG compression[7].In ad-dition,they proposed a downsampling technique that adapts downsampling factor,direction and DCT quantization step size to local image characteristics.This technique,in contrast to CADU,is tailored and only applicable to DCT-based JPEG image codec.Gan et al.also proposed undersampled boundary pre-and post-?lters to subdue blocking artifacts of DCT-based block codecs at low bit rates[8].

Apparently,the aforementioned previous works were largely motivated by the susceptibility of DCT-based block codecs to severe artifacts at low bit rates.However,in our view,for appli-cations of very tight bit budget,one should adopt wavelet-based codecs,such as JPEG2000,which are well known for their clear superiority over DCT-based codecs in both subjective and ob-jective quality at low bit rates.Now the question is whether it is also advantageous to downsample an image prior to wavelet-based compression at low bit rates.This paper will answer this question af?rmatively.In fact,according to our experimental results,the CADU technique coupled with JPEG2000consis-tently has higher PSNR than all existing techniques for bit rates below0.3bpp,and it achieves better visual quality than others for bit rates up to0.5bpp.

A precursor of this work is the method of coding quincunx down-sampled image[9].However,none of the compression standards accepts quincunx image as input.The encoder of the quincunx method was computationally expensive because it needed an adaptive directional lifting wavelet transform to code quincunx sample grid.Even with its high cost and standard noncompliance,the quincunx method yet has a slight worse compression performance than the new method of square lattice subsampling,as demonstrated in Section IV.

The paper is organized as follows.Section II describes the CADU encoder of uniform down-sampling with adaptive direc-tion pre?ltering.Section III presents the main contribution of this paper:a constrained least squares upconversion algorithm driven by a piecewise autoregressive image model.We report and discuss the experimental results in Section IV.

II.U NIFORM D OWN-S AMPLING W ITH A DAPTIVE

D IRECTIONAL P REFILTERING

Out of practical considerations,we make a more compact rep-resentation of an image by decimating every other row and every other column of the image.This simple approach has an oper-ational advantage that the down-sampled image remains a uni-form rectilinear grid of pixels and can readily be compressed by any of existing international image coding standards.To pre-vent the down-sampling process from causing aliasing artifacts, it seems necessary to low-pass pre?lter an input image to half of its maximum

frequency.However,on a second re?ection, one can do somewhat better.In areas of edges,the2-D spec-trum of the local image signal is not isotropic.Thus,we seek to perform adaptive sampling,within the uniform down-sampling framework,by judiciously smoothing the image with directional low-pass pre?ltering prior to down-sampling.

To this end,we design a family of2-D directional low-pass pre?lters under the criterion of preserving the maximum2-D bandwidth without the risk of aliasing.

Let

and

be the side lengths of the rectangular low-passed region of the 2-D?lter in the low-and high-frequency directions of an edge

554IEEE TRANSACTIONS ON IMAGE PROCESSING,VOL.18,NO.3,MARCH

2009

Fig.1.Block diagram of the proposed CADU image compression

system.

Fig.2.Directional ?lters of maximum passed region

w ( )

w ( )= .(a) =0;(b) =tan (1=2);(c) =( =4).

TABLE I

D ESIGN OF D IRECTIONAL L OW -P ASS 2-D P

REFILTERS

of

angle ,respectively.The maximum area of this low-passed

region without aliasing

is

.It is easy to show that there are only eight values

of (corresponding to three combinations

of

and values)to

achieve

while avoiding aliasing.These eight cases

are tabulated in Table I.Fig.2illustrates the above directional

low-pass ?lter design

for

,

,(the low-passed frequency range for other angles can be obtained by ro-tation).The spectra in the diagrams are those of the straight line of

angle .

In addition,the directional low-pass ?lter design serves two other purposes:1)most ef?cient packing of signal energy in presence of edges;2)preservation of subjective image quality for the edge is an important semantic construct.Moreover,as we will see in the next section,the use of low-pass pre?lters estab-lishes sample relations that play a central role in the decoding process of constrained least squares upconversion.

Many implementations of directional low-pass pre?lters are possible.For instance,the following directional low-pass pre-?lter can be

used:

(1)

where is the normalization factor to keep the ?lter in unit en-ergy,

and

is a window function (such as

the window function).The

parameters

and

are

(2)

In the directional pre?ltering step,the CADU encoder ?rst computes the gradient at the sampled position.If the amplitude of the gradient is below a threshold,the isotropic low-pass

?lter is applied.Otherwise,the gradient direction is quantized and

the corresponding

?lter

in Table I is selected and applied.Despite its simplicity,the CADU compression approach via uniform down-sampling is not inherently inferior to other image compression techniques in rate-distortion performance,as long as the target bit rate is below a threshold.The argument is based on the classical water-?lling principle in rate-distortion

theory.To encode a set

of

independent Gaussian random

variables

,,the rate-distor-tion bounds,when the total bit rate

being and

the total mean-squares distortion

being

,are given

by

(3)

超声诊断仪基本原理及其结构

江西中医学院计算机学院08生物医学工程2班黄月丹学号2 超声诊断仪原理及其基本结构 超声成像检查技术是指运用超声波的物理特性,通过高科技电子工程技术对超声波发射、接收、转换及电子计算机的快速分析处理和显像,从而对人体软组织的物理特性、形态结构与功能状态作出判断的一种非创性检查技术。 超声诊断技术的发展历程 20世纪50年代建立,70年代广泛发展应用的超声诊断技术,总的发展趋势是从静态向动态图像(快速成像)发展,从黑白向彩色图像过渡,从二维图像向三维图像迈进,从反射法向透射法探索,以求得到专一性、特异性的超声信号,达到定量化、特异性诊断的目的。80年代介入性超声逐渐普及,体腔探头和术中探头的应用扩大了诊断范围,也提高了诊断水平,90年代的血管内超声、三维成像、新型声学造影剂的应用使超声诊断又上了一个新台阶。 二.超声诊断仪的种类 (一) A型这是一种幅度调制超声诊断仪,把接收到的回声以波的振幅显示,振幅的高低代表回声的强弱,以波型形式出现,称为回声图,现已被B型超声取代,仅在眼科生物测量方面尚在应用,其优点是测量距离的精度高。(二) B型这是辉度调制型超声诊断仪,把接收到的回声,以光点显示,光点的灰度等级代表回声的强弱。通过扫

描电路,最后显示为断层图像,称为声像图。B型超声诊断仪由于探头和扫描电路的不同,显示的声像图有矩形、梯形和扇形。矩形声像图和梯形声像图用线阵探头实现,适用于浅表器官的诊断;扇形声像图用的探头有多种,机械扇扫探头、相控阵探头和凸阵探头均显示扇形声像图。前二种探头可由小的声窗窥见较宽的深部视野,适用于心脏诊断;后一种探头浅表与深部显示均宽广,适用于腹部诊断,有一种曲率半径小的凸阵探头,也可用小的声窗,窥见深部较宽的视野。 (三) M型 M型超声诊断仪是B型的一种变化,介于A型和B型之间,得到的是一维信息。在辉度调制的基础上,加上一个慢扫描电路,使辉度调制的一维回声信号,得到时间上的展开,形成曲线。用以观察心脏瓣膜活动等,现在M型超声已成为B型超声诊断仪中的一个功能部分不作为单独的仪器出售。(四) D型在二维图像上某点取样,获得多普勒频谱加以分析,获得血流动力学的信息,对心血管的诊断极为有用,所用探头与B型合用,只有连续波多普勒,需要用专用的探头。超声诊断仪兼有B型功能和D型功能者称双功超声诊断仪。(五) 彩色多普勒超声诊断仪具有彩色血流图功能,并覆盖在二维声像图上,可显示脏器和器官内血管的分布、走向,并借此能方便地采样,获得多普勒频谱,测得血流的多项重要的血流动力学参数,供诊断之用。彩色多普勒超声诊断仪一般均兼有B型、M型、D型和彩色血流图功能。(六) 三维超声诊断仪三维超声是建立在二维基础上,在彩色多普勒超声诊断仪的基础上,配上数据采集装置,再加上三维重建软件,该仪器即有三维显示功能。(七) C型C型超声仪也是辉度调制型的一种,与B型不同的是其显示层面与探测面呈同等深度。超声诊断仪基本原理

光电成像系统的分辨率鉴定与测量技术

光电成像系统的分辨率鉴定与测量技术 摘要:论述了光电成像系统中广泛使用的分辨率指标及分类,对空间分辨率模拟度量法的原理和测量方法进行了论述和分析。通过研究指出用空间分辨率指标来描述成像系统的质量,具有较好的直观性和归一性。由于单一的空间分辨率测量指标还不可能给出总的图像系统的性能,仅仅基于分辨率指标的图像评估不可能同时保证系统灵敏度设计的技术要求。因此,结合模拟度量法研究光电成像系统的分辨率测量法,给出成像分辨率测量准则。 关键词:MTF;SRF;空间分辨率;DAS;GRD 中图分类号:TP29文献标识码:A 文章编号:1004-373X(2010)01-177-03 Resolution Identification and Measuring Technique of Photoelectric Image System ZHANG Bin,LI Zhaohui (Chinese Flight Test Establishment,Xi′an,710089,China) Abstract:Index and classification of resolution which are widely used in the photoelectric image system is discussed with analysis of the principle and method of the simulated measurement of spatial resolution.The investigation shows that

the index of spatial resolution which describes quality of the image-forming system is more direct and unitary than other methods.However,the single spatial resolution can not show the capability of the whole image system.Besides,the evaluation which it is only based on the index of spatial resolution can not ensure the designed technical requirement of the system sensitivity.Therefore,on the basis of the resolution measuring method of the photoelectric image system,a measuring criterion of the imaging resolution is obtained. Keywords:MTF;SRF;spatial resolution;DAS;GRD 0 引言 物理系统中对分辨率指标的使用由来已久,它是确定成像系统性能指标的基本要素,尤其是用分辨率作为衡量图像质量的指标之一,人们会因此认为具有较高分辨率的系统具有较好的图像质量[1]。一般情况下,对于类似于系统设计这样的问题确实如此(例如,将两个EMUX系统相比),其MTF(调制传递函数)具有相同的函数形式。 分辨率有四类不同内容[2]:时间分辨率(以时间分类事件的能力);灰度分辨率(由A/D变换器设计、噪声低限、或监视器性能指标决定);谱分辨率;空间分辨率。

CCD相机成像分辨率自动测试的过程与方法介绍

CCD相机成像分辨率自动测试的过程与方法介绍 引言 目前传输型CCD相机已取代传统胶片相机成为主流摄影设备,然而各生产厂家对相机成像分辨率这一核心指标的测量还基本采用基于人工判读的测试方法。 人工判读测试分辨率,对胶片相机而言简单、方便,但由于不同人眼的视觉灵敏度不同以及检测条件的差异,因此难免引入不同程度的主观误差,时常难以达成统一的测量结果,从而影响了测试精度。 对于CCD 相机,可利用其对特定目标生成的数字影像,通过实施高效的数据分析处理技术,自动实现对相机分辨率量化测试,从而客观判定相机成像质量。 1 理论分析 影响CCD相机成像分辨率的因素主要包括:光学系统、CCD器件及相应电路处理系统等。其中光学系统可利用干涉检测法或传递函数等对其像质进行测试,从而客观地获取相应的分辨率量化结果;CCD器件本身的理论极限分辨率可以根据其像元尺寸直接计算求得;对于电路处理系统,在理想情况下其对图像分辨率测试方面的影响可忽略不计,在此暂不予以考虑。综合上述因素,CCD 相机整机理想情况下的分辨率N 可由下式计算求得: 式中:N光为光学系统分辨率;NCCD 为CCD器件的分辨率。 虽然上述计算可以估算出CCD相机整机的理论分辨率,但由于存在整机装配误差、系统控制误差以及依靠人工判读测试带来的主观不确定性,经常难以准确反映相机最终成像水平,因此需要在CCD 相机整机检测时对分辨率指标实施精确量化测试,从而客观综合反映CCD相机整机成像质量。 为此,本文提出基于光栅目标影像对比度分析的分辨率自动测试方法。该方法是将CCD 相机整体作为光能量信息传递系统,根据系统传递函数测试原理,按照正弦级数展开的定义,将矩形分布函数展开成不同频率正弦分布的叠加,则对比度传递函数可表示为:

超声成像基础原理以及心脏超声

超声成像 学习要求:掌握超声成像的基本原理(超声、超声的物理特性及其应用)、超声图像的特点了解超声波的产生、超声成像、超声检查技术与设备,超声诊断的方法学目的:理解超声诊断的临床应用 超声成像的定义:利用超声波的物理特性和人体器官组织声学特征相互作用后所产生的信息,经信息处理形成图像的成像技术,借此进行疾病诊断的检查方法。 一、超声波的物理特性(1): 波可分为:电磁波(包括可见光、无线电波、X线)和机械波(包括声波、水波、地震波)声波:20~20000 Hz 超声波:>20000 Hz 医用超声波:2.5~10 MHz 二、超声波的物理特征(2) 1.超声波的物理量(波长、频率、传播速度)及其关系: 物理量: 频率(f) : Hz 声速(c) : m /s 或cm/s 波长(λ) : m 介质密度(ρ) : g/cm3 声阻抗(Z):Z=ρ×c(g/cm2.s) 关系: c2=K / ρ即声速取决于波长和频率, 并与介质中的弹性(K) 和密度(ρ) 密切相关c=f ×λ即同一介质中传播(C确定),频率越高则波长越短 传播速度: 固体>液体>气体 2.束射性或指向性(超声波的直线传播) 其方向性与超声频率、声源直径及后者与波长的比值有关 扩散角越小,方向性越好 3.反射:超声在均质性介质传播中不出现反射 反射条件: ①介质声阻抗差>0.1% ②界面大于波长 声阻抗=介质密度与速度的乘积 4.散射

超声波在介质中传播如遇不规则的小界面, 或界面小于波长时,则发生散射 5.衰减: 超声波在介质中传播由于介质吸收(声能转化为热) 、反射、散射等原因,其振幅与强度逐渐降低,这种现象称为衰减。(振幅与强度的减小) 6.多普勒效应: 声束在介质中传播时,如遇到运动的反射界面,其反射的超声波频率随界面运动的情况而发生改变的现象 三、超声波的产生: 1、压电晶片(换能器) 2、压电效应:逆压电效应(电能转变为声能) 正压电效应 四、超声成象基本原理 1、器官、组织中各种界面对超声波的不同反射和/或散射是构成图象的基础。 2、仪器将接收到的含有各种声学信息的回声,经过处理,在显示器上显示为波形、曲线、图象 五、超声诊断的种类 1、A型---A mplitude 以波的形式显示出来,为幅度调制型 2、M型---M otion echocardiography 是B型超声中的一种特殊显示方式 3、B型---B rightness 以光点的形式显示出来,为辉度调制型 扫查连续, 由点, 线而扫描出脏器的解剖切面, 是二维空间显示, 又称二维法 4、D型---D oppler ( pw、cw、color doppler) 彩色多普勒血流显像CDFI(color Doppler flow imaging): 将二维彩色血流信号重叠到二维B型扫描或M型扫描图上,实现解剖结构与血流状态两种图像结合的实时显像 用红, 黄, 蓝三种基本颜色编码,显示不同血流方向 颜色的辉度与血流速度成正比 彩色多普勒血流显像不仅能清楚的显示心脏大血管的形态结构和活动情况,而且能直观和形象地显示心内血流的方向、速度、范围、有无血流紊乱及异常通路等 ——故有人称之为非损伤性心血管造影法。 六、超声图像特点:

X射线实时成像分辨率

X射线实时成像系统分辨率及其影响因素 摘要: 概述了X射线实时成像系统的基本配置和反映系统质量特性的调制传递函数以及提高X射线实时成像系统分辨率的基本方法。 关键词:系统分辨率质量特性调制传递函数 The Resolution and Influencing Factor in X-Ray Real Time Image System Zeng Xiangzhao (Nanhai Yuehai Steel Products Co.,Ltd Guangdong 528247) Abstract: This article introduced the basic configure and the modulating transfer function which reflect the systemic quality speciality in x-ray real timeimage system, and introduced the basic technique for enhance the systemic resolution in X-Ray real time image system Keywords:System Resolution Quality speciality The modulating transfer function 1 X射线实时成像系统 X射线实时成像检测技术作为一种新兴的无损检测技术,已进入工业产品检测的实际应用领域。与其他检测技术一样,X射线实时成像检测技术需要一套设备(硬件与软件)作为支撑,构成一个完整的检测系统,简称X射线实时成像系统。X射线实时成像系统使用X射线机或加速器等作为射线源,X射线透过后被检测物体后衰减,由射线接收/转换装置接收并转换成模拟信号或数字信号,利用半导体传感技术、计算机图像处理技术和信息处理技术,将检测图像直接显示在显示器屏幕上,应用计算机程序进行评定,然后将图像数据保存到储存介质上。X射线实时成像系统可用金属焊缝、金属或非金属器件的无损检测。 2 X射线实时成像系统的基本配置及影响因素 X射线实时成像系统主要由X射线机、X射线接收转换装置、数字图像处理单元、图像显示单元、图像储存单元及检测工装等组成。 2.1 X射线机 根据被检测工件的材质和厚度范围选择X射线机的能量范围,并应留有一定的的能量储备。对于要求连续检测的作业方式,宜选择直流恒压强制冷却X射线机。X射线管的焦点尺寸对检测图像质量有较大的影响,小焦点能够提高系统分辨率,因此,应尽可能选用小焦点X射线管。 目前探伤机厂能够提供的小焦点X射线探伤机是:160 kV恒压式X射线系统,焦点尺寸≤ 0.4mm×0.4mm;225 kV恒压式X射线系统,焦点尺寸≤0.8mm×0.8mm;320 kV恒压式X射线系统,焦点尺寸≤1.2mm×1.2mm;450 kV恒压式X射线系统,焦点尺寸≤1.8mm×1.8mm。对焦点的要求也不宜过小,如果焦点过小且冷却不好,焦点容易"烧坏"。 2.2 X射线接收转换装置 X射线接收转换装置的作用是将不可见的X光转换为可见光,它可以是图像增强器或成像面板或者线性扫描器等射线敏感器件。X射线接收转换装置的分辨率应不小于3.0LP/mm。 X射线接收转换装置子系统又称为图像成像系统,按目前成像的技术水平可分为两种。一种是以图像增强器为主的传统成像器系统。图像增强器为一种真空管,射线输入屏由较薄的铝或钛材料制成,屏的基层涂有钠(Na)-碘化铯(CsI)作为输入闪烁体(CsI∶Na),它能够将不可光的X 光图像转换为可见光图像,再经过光电阴极板的作用将可见光图像转换为相应的电子束,电子束在高电压作用下加速并聚焦于荧光输出屏(ZnCdS:Ag闪烁体材料),从而形成可视的检测图像。在输出屏后端配有聚焦光学镜头和CCD(charge-coupled device电荷耦合器件)摄像机,将可视图像的模拟信号采集输入图像采集卡进行A/D转换,再输入计算机进行图像处理。当前可供选用的图像增强器按输入屏直径有Φ225mm(9″)、Φ150mm(6″)、Φ100mm(4″) 三种;Φ225mm(9″)图像增强器直径较大,视野宽阔,一次检测长度较大,但清晰度较低,价格较高;Φ100mm(4″)图像增强器直径较小,重量较轻,便于携带式作业,且清晰度较高,但视野较狭小,一次检测长

超声诊断仪基本原理和结构

江西中医学院计算机学院08生物医学工程2班黄月丹学号5047 超声诊断仪原理及其基本结构 超声成像检查技术是指运用超声波的物理特性,通过高科技电子工程技术对超声波发射、接收、转换及电子计算机的快速分析处理和显像,从而对人体软组织的物理特性、形态结构与功能状态作出判断的一种非创性检查技术。 超声诊断技术的发展历程 20世纪50年代建立,70年代广泛发展应用的超声诊断技术,总的发展趋势是从静态向动态图像(快速成像)发展,从黑白向彩色图像过渡,从二维图像向三维图像迈进,从反射法向透射法探索,以求得到专一性、特异性的超声信号,达到定量化、特异性诊断的目的。80年代介入性超声逐渐普及,体腔探头和术中探头的应用扩大了诊断范围,也提高了诊断水平,90年代的血管内超声、三维成像、新型声学造影剂的应用使超声诊断又上了一个新台阶。 二.超声诊断仪的种类 (一) A型这是一种幅度调制超声诊断仪,把接收到的回声以波的振幅显示,振幅的高低代表回声的强弱,以波型形式出现,称为回声图,现已被B型超声取代,仅在眼科生物测量方面尚在应用,其优点是测量距离的精度高。(二) B型这是辉度调制型超声诊断仪,把接收到的回声,以光点显示,光点的灰度等级代表回声的强弱。通过扫

描电路,最后显示为断层图像,称为声像图。B型超声诊断仪由于探头和扫描电路的不同,显示的声像图有矩形、梯形和扇形。矩形声像图和梯形声像图用线阵探头实现,适用于浅表器官的诊断;扇形声像图用的探头有多种,机械扇扫探头、相控阵探头和凸阵探头均显示扇形声像图。前二种探头可由小的声窗窥见较宽的深部视野,适用于心脏诊断;后一种探头浅表与深部显示均宽广,适用于腹部诊断,有一种曲率半径小的凸阵探头,也可用小的声窗,窥见深部较宽的视野。 (三) M型 M型超声诊断仪是B型的一种变化,介于A型和B型之间,得到的是一维信息。在辉度调制的基础上,加上一个慢扫描电路,使辉度调制的一维回声信号,得到时间上的展开,形成曲线。用以观察心脏瓣膜活动等,现在M型超声已成为B型超声诊断仪中的一个功能部分不作为单独的仪器出售。(四) D型在二维图像上某点取样,获得多普勒频谱加以分析,获得血流动力学的信息,对心血管的诊断极为有用,所用探头与B型合用,只有连续波多普勒,需要用专用的探头。超声诊断仪兼有B型功能和D型功能者称双功超声诊断仪。(五) 彩色多普勒超声诊断仪具有彩色血流图功能,并覆盖在二维声像图上,可显示脏器和器官内血管的分布、走向,并借此能方便地采样,获得多普勒频谱,测得血流的多项重要的血流动力学参数,供诊断之用。彩色多普勒超声诊断仪一般均兼有B型、M型、D型和彩色血流图功能。(六) 三维超声诊断仪三维超声是建立在二维基础上,在彩色多普勒超声诊断仪的基础上,配上数据采集装置,再加上三维重建软件,该仪器即有三维显示功能。(七) C型C型超声仪也是辉度调制型的一种,与B型不同的是其显示层面与探测面呈同等深度。超声诊断仪基本原理

超声成像原理

第一章超声成像原理和妇产超声诊断临床基础 第一节超声成像原理 一、超声波的概念和基本特性 (一)超声波的概念频率在2万赫兹以上的机械振动波,称为超声波(ultrasonic wave),简称超声(ultrasound)。能够传递超声波的物质,称为传声介质,它具有质量和弹性,包括各种气体、液体和固体;传声介质有均匀的、不均匀的;有各向同性的、各向异性的等。超声波在传声介质中的传播特点是具有明确指向性的束状传播,这种声波能够成束地发射并用于定向扫查人体组织。 (二)超声波的产生医用高频超声波是由超声诊断仪上的压电换能器产生的,这种换能器又称为探头,能将电能转换为超声能,发射超声波,同时,它也能接受返回的超声波并把它转换成电信号。探头具有发射和接受超声两种功能。常用的探头分为线阵型、扇型、凸阵型,探头的类型不同,发射的超声束形状和大小各不相同,而各种探头根据探查部位的不同被设计成不同的形状。见图1-1-1。 图1-1-1 探头示意 (三)超声波的基本物理量 1.频率(f):是指单位时间内质点振动的次数。单位是赫兹(Hz)、千赫(KHz)、兆赫(MHz)。超声的频率在20KHz以上,而医学诊断用超声的频率一般在兆赫级,称为高频超声波,常用频率范围2~10兆赫。频率越高,波的纵向分辨力越好。周期(T)则是一个完整的波通过某点所需的时间。有f·T = 1 。 2.波长(λ):表示在均匀介质中的单频声波行波振动一个周期时间内所传播的距离,也就是一个波周期在空间里的长度。波的纵向分辨力的极限是半波长,因此了解人体软组织中传

导的超声波长有助于估计超声波分辨病灶大小的能力。 3.声速(C):是指声波在介质中传播的速度。声速是由弹性介质的特性决定的,不同介质的声速是不同的。人体各种软组织之间声速的差异很小,约5%左右,所以在各种超声诊断仪器检测人体脏器时,假设各种软组织的声速是相等的,即采用了人体软组织平均声速的概念。目前,较多采用人体软组织平均声速的数值是1540m/s。实际上人体不同软组织脏器及体液的声速是有差别的,因此声像图上显示的目标,无论是脏器或病灶,其位置及大小与实际的结构相比,都存在误差,但不致影响诊断结论,一般可忽略 声速C、波长λ、频率f或周期T之间的关系符合 4.声强(sound intensity):当声波在介质中传播时,声波的能量从介质的一个体积元通过邻近的体积元向远处传播。 声强是指超声波在介质中传播时,单位时间内通过垂直于传播方向的单位面积的平均能量。声强的物理意义为单位时间内在介质中传递的超声能量,或称超声功率。声强小时超声波对人体无害,声强超过一定限度,则可能对人体产生伤害,目前规定临床超声诊断仪安全剂量标准为平均声强小于10mW/cm2。(四)超声波的传播 1. 声特性阻抗(acoustic characteristic impedance):声特性阻抗(Z)定义为平面自由行波在介质中某一点处的声压(p)与质点速度(u)的比值。在无衰减的平面波的情况下,声特性阻抗等于介质的密度(ρ)与声速(C)的乘积。 2. 声特性阻抗差与声学界面:两种介质的声特性阻抗差大于1‰时,它们的接触面即可构成声学界面。入射的超声波遇声学界面时可发生反射和折射等物理现象。人体软组织及脏器结构声特性阻抗的差异构成大小疏密不等、排列各异的声学界面,是超声波分辨组织结构的声学基础。 3. 声波的界面反射与折射:超声入射到声学界面时引起返回的过程,称为声反射(acoustic reflection)。射向声学界面的入射角等于其反射角。而声波穿过介质之间的界面,进入另一种介质中继续传播的现象,称为声透射(acoustic transmission)。当超声的入射方向不

突破衍射极限的超高分辨率成像技术发展(修改)

突破衍射极限的超高分辨率成像技术发展(修改) -标准化文件发布号:(9456-EUATWK-MWUB-WUNN-INNUL-DDQTY-KII

结课论文 题目突破衍射极限的超高分辨率成像的技术进展 学生姓名 学号 学院 专业 班级 二〇一五年十二月

一引言 1.1选题意义 光学显微成像具有极为悠久的历史,但一直以来,光学成像一直受到衍射极限的限制而分辨率无法突破200 nm。后来虽然有了电子显微镜、核磁共振显像、x光衍射仪等微观观测或者显像设备,但是使用光学显微镜可以在活体状态下观察生命体使得其在生物、医学观察方面仍有巨大优势。值得庆贺的是近年来,超高分辨率显微技术的发展使得光学显微成像分辨率达到了20 nm以下。其中德国科学家Stefan Hell、美国科学家Eric Betzig和William Moerner因其在超高分辨率显微技术方面的突出贡献获得了2014年的诺贝尔化学奖。在这篇文章中,我们就简要介绍一下超高分辨率显微技术的发展和应用,并对诸位大师致以敬意。 1.2技术 指标 显微技术成像优劣一般通过X-Y平面分辨率与Z 轴分辨率大小来判定,分辨率越高数值越 小。下表是各种显微成像技术的分辨率指标。

STED显微技术50-70 STED+4技术50 50 PALM技术20 30 3D STORM技术20-30 50-60 dSTORM技术30 50 2D SSIM技术50 3D SSIM技术100 200 电子显微镜 X光衍射仪

二衍射极限 衍射极限 我们能看到什么看到多小的范围看得有多清楚几百年来,依靠不断进步的科学手段,微观世界正一层层揭开面纱,让人们可以看得越来越“小”,进而可以进行研究。 人的肉眼能分辨毫米尺度的物体,再小,就要借助工具。1665年,英国科学家罗伯特·虎克制造了第一台用于科学研究的光学显微镜,用它观察薄薄的软木塞切片。虎克看到了残存的植物细胞壁,它们一个个像小房间一样紧挨在一起,这就是“细胞”一词的由来。 此后,显微镜制造和显微观察技术的迅速发展,帮助科学家第一次发现了细菌和微生物。那么,光学显微镜是否可以无止境地“放大”下去,让我们想看到多小就能看到多小科学家为此做了很多尝试,最终发现,存在一道法逾越的“墙”—衍射极限。 1873年,德国科学家阿贝提出了衍射极限理论:光是一种电磁波,由于存在衍射,一个被观测的点经过光学系统成像后,不可能得到理想的点,而是一个衍射像,每个物点就像一个弥散的斑,如果这两个点靠得很近,弥散斑就叠加在一起,我们看到的就是一团模糊的图像。 阿贝提出,分辨率的极限近似于入射光波长的二分之一(d=λ/2)。可见光的波长通常在380~780纳米之间,根据衍射极限公式,光学显微镜的分辨率极限就在200纳米(微米)左右。如果物体小于微米,你仍旧看到的是一个模糊的光斑。这就是很长一段时间内,光学显微镜的分辨极限——衍射极限。

X射线实时成像系统分辨率及其影响因素

X射线实时成像系统分辨率及其影响因素 1 X射线实时成像系统 X射线实时成像检测技术作为一种新兴的无损检测技术,已进入工业产品检测的实际应用领域。与其他检测技术一样,X射线实时成像检测技术需要一套设备(硬件与软件)作为支撑,构成一个完整的检测系统,简称X射线实时成像系统。X射线实时成像系统使用X射线机或加速器等作为射线源,X射线透过后被检测物体后衰减,由射线接收/转换装置接收并转换成模拟信号或数字信号,利用半导体传感技术、计算机图像处理技术和信息处理技术,将检测图像直接显示在显示器屏幕上,应用计算机程序进行评定,然后将图像数据保存到储存介质上。X射线实时成像系统可用金属焊缝、金属或非金属器件的无损检测。 2 X射线实时成像系统的基本配置及影响因素 X射线实时成像系统主要由X射线机、X射线接收转换装置、数字图像处理单元、图像显示单元、图像储存单元及 检测工装等组成。 2.1 X射线机 根据被检测工件的材质和厚度范围选择X射线机的能量范围,并应留有一定的的能量储备。对于要求连续检测的作业方式,宜选择直流恒压强制冷却X 射线机。X射线管的焦点尺寸对检测图像质量有较大的影响,小焦点能够提高系统分辨率,因此,应尽可能选用小焦点X射线管。 2.2 X射线接收转换装置 X射线接收转换装置的作用是将不可见的X光转换为可见光,它可以是图像增强器或成像面板或者线性扫描器等射线敏感器件。X射线接收转换装置的分辨率应不小于3.0LP/mm。 2.3 图像处理单元 图像处理单元应具有图像数据采集和处理功能。图像数据采集方式可以是图像采集卡或其它数字图像合成装置。图像采集分辨率应不低于768×576像素,且保证水平方向分辨率与垂直方向分辨率之比为4∶3;动态范围即灰度等级应不小于256级。 2.4 图像处理软件 图像处理软件应具有降噪、亮度对比度增强、边缘增强等基本功能。图像处

提高超声成像的分辨率

552IEEE TRANSACTIONS ON IMAGE PROCESSING,VOL.18,NO.3,MARCH2009 Low Bit-Rate Image Compression via Adaptive Down-Sampling and Constrained Least Squares Upconversion Xiaolin Wu,Senior Member,IEEE,Xiangjun Zhang,and Xiaohan Wang Abstract—Recently,many researchers started to challenge a long-standing practice of digital photography:oversampling followed by compression and pursuing more intelligent sparse sampling techniques.In this paper,we propose a practical ap-proach of uniform down sampling in image space and yet making the sampling adaptive by spatially varying,directional low-pass pre?ltering.The resulting down-sampled pre?ltered image re-mains a conventional square sample grid,and,thus,it can be compressed and transmitted without any change to current image coding standards and systems.The decoder?rst decompresses the low-resolution image and then upconverts it to the original resolu-tion in a constrained least squares restoration process,using a2-D piecewise autoregressive model and the knowledge of directional low-pass pre?ltering.The proposed compression approach of collaborative adaptive down-sampling and upconversion(CADU) outperforms JPEG2000in PSNR measure at low to medium bit rates and achieves superior visual quality,as well.The superior low bit-rate performance of the CADU approach seems to suggest that oversampling not only wastes hardware resources and energy, and it could be counterproductive to image quality given a tight bit budget. Index Terms—Autoregressive modeling,compression standards, image restoration,image upconversion,low bit-rate image com-pression,sampling,subjective image quality. I.I NTRODUCTION O VER the past half century,the prevailing engineering practice of image/video compression,as exempli?ed by all existing image and video compression standards,is to start with a dense2-D sample grid of https://www.doczj.com/doc/882066110.html,pression is done by transforming the spatial image signal into a space(e.g.,spaces of Fourier or wavelet bases)in which the image has a sparse rep-resentation and by entropy coding of transform coef?cients.Re-cently,researchers in the emerging?eld of compressive sensing challenged the wisdom of what they called“oversampling fol-lowed massive dumping”approach.They showed,quite sur-prisingly,it is possible,at least theoretically,to obtain compact signal representation by a greatly reduced number of random samples[1]. Manuscript received March12,2008;revised November05,2008.Current version published February11,2009.The associate editor coordinating the re-view of this manuscript and approving it for publication was Prof.Bruno Car-pentieri. The authors are with the Department of Electrical and Computer Engineering, McMaster University,Hamilton,ON L8S4K1Canada(e-mail:xwu@ece.mc-master.ca;zhangxj@grads.ece.mcmaster.ca;wangx28@mcmaster.ca). Color versions of one or more of the?gures in this paper are available online at https://www.doczj.com/doc/882066110.html,. Digital Object Identi?er10.1109/TIP.2008.2010638 In this paper,we investigate the problem of compact image representation in an approach of sparse sampling in the spa-tial domain.The fact that most natural images have an expo-nentially decaying power spectrum suggests the possibility of interpolation-based compact representation of images.A typ-ical scene contains predominantly smooth regions that can be satisfactorily interpolated from a sparsely sampled low-resolu-tion image.The dif?culty is with the reconstruction of high fre-quency contents.Of particular importance is faithful reconstruc-tion of edges without large phase errors,which is detrimental to perceptual quality of a decoded image. To answer the above challenge,we develop a new image com-pression methodology of collaborative adaptive down-sampling and upconversion(CADU).The CADU approach puts an em-phasis on edge reconstruction for the perceptual importance of edges and also to exploit the anisotropy of edge spectrum.In the CADU encoder design,we choose not to perform uneven irregular down sampling of an input image according to local spatial or frequency characteristics.Instead,we stick to con-ventional square pixel grid by uniform spatial down sampling of the image.Consequently,the resulting low-resolution image can be readily compressed by any of existing image compres-sion techniques,such as DCT-based JPEG and JPEG2000.Yet the simple uniform down-sampling scheme is made adaptive by a directional low-pass pre?ltering step prior to down-sampling. The other purpose of this preprocessing is to induce a mecha-nism of collaboration between the spatial uniform down-sam-pling process at the encoder and optimal upconversion process at the decoder. The CADU decoder?rst decompresses the low-resolution image and then upconverts it to the original resolution in con-strained least squares restoration process,using a2-D piece-wise autoregressive model(PAR)and by reversing the direc-tional low-pass pre?ltering operation of the encoder.Two-di-mensional autoregressive modeling was a known effective tech-nique of predictive image coding[2],[3].For the CADU de-coder,the PAR model plays a role of adaptive noncausal pre-dictor.What is novel and unique of the CADU approach is that the predictor is only used at the decoder side,and the noncausal predictive decoding is performed in collaboration with the pre-?ltering of the encoder. Fig.1presents a block diagram of the proposed CADU image compression system,summarizing the above ideas and depicting the collaboration between the encoder and decoder. The CADU image compression technique,although oper-ating on down-sampled images,obtains some of the best PSNR 1057-7149/$25.00?2009IEEE

(完整word版)计算成像

计算成像 1.为何要研究计算成像? 计算成像能够实现传统成像无法完成的任务,例如:去除运动模糊、超分辨率重建等。 2.用计算成像的方法怎样提高图像分辨率? 增加空间分辨率最直接的解决方法就是通过传感器制造技术减少像素尺寸(例如增加每单元面积的像素数量)。然而,随着像素尺寸的减少,光通量也随之减少,它所产生的散粒噪声使得图像质量严重恶化。不受散粒噪声的影响而减少像素的尺寸有一个极限,对于0.35微米的CMOS处理器,像素的理想极限尺寸大约是40平方微米。当前的图像传感器技术大多能达到这个水平。 另外一个增加空间分辨率的方法是增加芯片的尺寸,从而增加图像的容量。因为很难提高大容量的耦合转换率,因此这种方法一般不认为是有效的。在许多高分辨率图像的商业应用领域,高精度光学和图像传感器的高价格也是一个必须考虑的重要因素。因此,有必要采用一种新的方法来增加空间分辨率,从而克服传感器和光学制造技术的限制。 (1)超分辨率重建 图像超分辨率是指由一幅低分辨率图像或图像序列恢复出高分辨率图像,高分辨率图像意味着图像具有高像素密度,可以提供更多的细节,这些细节往往在应用中起到关键作用。图像超分辨率技术分为超分辨率复原和超分辨率重建。超分辨率重构的基本过程为:先进行图像退化分析,然后进行图像的配准,最后根据配准的信息对图像进行重构。目前,图像超分辨率研究可分为3个主要范畴:基于插值、基于重建和基于学习的方法。具体方法有:规整化重建方法,均匀空间样本内插方法,迭代反投影方法(IBP),集合理论重建方法(凸集投影POCS),统计重建方法(最大后验概率MAP和最大似然估计ML),混合ML/MAP/POCS 方法,自适应滤波/维纳滤波/卡尔曼滤波方法,确定性重建方法基于学习和模式识别的方法。 超分辨率重建,即通过硬件或软件的方法提高原有图像的分辨率,通过一系列低分辨率的图像来得到一幅高分辨率的图像。超分辨率重建的核心思想就是用时间带宽(获取同一场景的多帧图像序列)换取空间分辨率,实现时间分辨率向空间分辨率的转换。在基于超分辨率重建的空间分辨率增强技术中,其基本前提是通过同一场景可以获取多幅低分辨率细节图像。在超分辨率重建中,典型地认为低分辨率图像代表了同一场景的不同侧面,也就是说低分辨率图像是基于亚像素精度的平移亚采样。如果仅仅是整数单位的像素平移,那么每幅图像中都包含了相同的信息,这样就不能为高分辨率图像的复原提供新的信息。如果每幅低分辨率图像彼此之间都是不同的亚像素平移,那么它们彼此之间就不会相互包含,在这种情况下,每一幅低分辨率图像都会为高分辨率图像的复原提供一些不同的信息。为了得到同一场景的不同侧面,必须通过一帧接一帧的多场景或者视频序列的相关的场景运动。我们可以通过一台照相机的多次拍摄或者在不同地点的多台照相机获取多个场景,例如在轨道卫星一类可控制的图像应用中,这种场景运动是能够实现的;对于局部对象移动或者震荡一类的不可控制的图像应用也是同样能实现的。如果这些场景运动是已知的或者是在亚像素精度范围了可估计的,同时如果我们能够合成这些高分辨率图像,那么超分辨率重建图像复原是可以实现的。

相关主题
文本预览
相关文档 最新文档