图像科学综述-外文翻译-外文文献-英文文献.doc

《图像科学综述-外文翻译-外文文献-英文文献.doc》由会员分享，可在线阅读，更多相关《图像科学综述-外文翻译-外文文献-英文文献.doc（9页珍藏版）》请在得力文库 - 分享文档赚钱的网站上搜索。

1、附录图像科学综述近几年来，图像处理与识别技术得到了迅速的发展，现在人们己充分认识到图像处理和识别技术是认识世界、改造世界的重要手段。目前它己应用于许多领域，成为2l世纪信息时代的一门重要的高新科学技术。1.图像处理与识别技术概述图像就是用各种观测系统以不同形式和手段观测客观世界而获得的，可以直接或间接作用于人眼而产生视知觉的实体。科学研究和统计表明，人类从外界获得的信息约有75%来自于视觉系统，也就是说，人类的大部分信息都是从图像中获得的。图像处理是人类视觉延伸的重要手段，可以便人们看到任意波长上所测得的图像。例如，借助伽马相机、x光机，人们可以看到红外和超声图像：借助CT可看到物体内部的断层

2、图像；借助相应工具可看到立体图像和剖视图像。1964年，美国在太空探索中拍回了大量月球照片，但是由于种种环境因素的影响，这些照片是非常不清晰的，为此，美国喷射推进实验室(JPL)使用计算机对图像进行处理，使照片中的重要信息得以清晰再现。这是这门技术发展的重要里程碑。此后，图像处理技术在空间研究方面得到广泛的应用。总体来说，图像处理技术的发展大致经历了初创期、发展期、普及期和实用化期4个阶段。初创期开始于20世纪60年代，当时的图像采用像素型光栅进行扫描显示，大多采用巾、大型机对其进行处理。在这一时期，由于图像存储成本高，处理设备造价高，因而其应用面很窄。20世纪70年代进入了发展期，开始大量采

3、用中、小型机进行处理，图像处理也逐渐改用光栅扫描显示方式，特别是出现了CT和卫星遥感图像，对图像处理技术的发展起到了很好的促进作用。到了20世纪80年代，图像处理技术进入普及期，此时购微机已经能够担当起图形图像处理的任务。VLSL的出现更使得处理速度大大提高，其造价也进一步降低，极大地促进了图形图像系统的普及和应用。20世纪90年代是图像技术的实用化时期，图像处理的信息量巨大，对处理速度的要求极高。21世纪的图像技术要向高质量化方面发展，主要体现在以下几点：高分辨率、高速度，图像处理技术发展的最终目标是要实现图像的实时处理，这在移动目标的生成、识别和跟踪上有着重要意义：立体化，立体化所包括的信

4、息最为完整和丰富，数字全息技术将有利于达到这个目的；智能化，其目的是实现图像的智能生成、处理、识别和理解。2.图像处理与识别技术的应用领域目前，图像处理与识别技术的主要应用领域有生物医学、文件处理、工业检测、机器人视觉、货物检测、邮政编码、金融、公安、银行、机械、交通、电子商务和多媒体网络通信等领域。3.图像处理与识别数字图像处理和识别学科所涉及的知识非常广泛，具体的方法种类繁多，应用也极为普遍，但从学科研究内容上可以分为以下几个方面：图像数字化、图像变换、图像增强、图像分割、图像分析。4.图像识别技术图像识别是近20年来发展起来的一门新型技术科学，它以研究某些对象或过程（统称图像）的分类与描

5、述为主要内容。图像识别所研究的领域十分广泛，它可以是医学图像中的癌细胞识别；机械加工中零部件的识别、分类；可以是认遥感图片中辨别农作物、森林、湖泊和军事设施，以及判断农作物的长势，预测收获量等；可以是自导引小车中的路径识别；邮政系统中自动分拣信函；交通管制、识别违章行驶的汽车牌照；银行的支票识别、身份证识别等。上述都是图像识别研究的课题。总体来说所研究的问题，主要是分类问题。5.图像处理在研究图像时，首先要对获得的图像信息进行预处理（前处理）以滤去干扰、噪声，作几何、彩色校正等。这样可提高信噪比；有时由于信息微弱，无法辨识，还得进行增强处理。增强的作用，在于提供一个满足一定要求的图像，或对图像

6、进行变换，以便人或计算机分析。并且为了从图像中找到需要识别的东西，还得对图像进行分割，也就是进行定位和分离，以分出不同的物体。为了给观察者以清晰的图像，还要对图像进行政善，即进行复原处理，它是把已经退化了图像加以重建或恢复的过程，以便改进图像的保真度。在实际处理中，由于图像信息量非常大，在存储及传送时，还要对图像信息进行压缩。上述工作必须用计算机进行，因而要进行编码等工作。编码的作用，是用最少数量的编码位(亦称比特)，表示单色和彩色图像，以便更有效地传输和存储。以上所述都属图像处理的范畴。因此，图像处理包括图像编码、图像增强、图像压缩、图像复原、图像分割等。对图像处理环节来说，输入是图像，输出

7、也是图像。由图像处理的内容可见，图像处理的目的主要在于解决两个问题：一是判断图像中有无需要的信息：另一是确定这些信息是什么。6.图像理解所谓图像理解是一个总称。上述图像处理及图像识别的最终目的，就在于对图像作描述和解释，以便最终理解它是什么图像。所以它是在图像处理及图像识别的基础上，再根据分类作结构句法分析，去描述图像和解释图像。因而图像理解包括图像处理、图像识别和结构分析。对理解部分米说，输入是图像，输出则是图像的描述与解释。7.图像识别与图像处理及图像理解的关系上面说过，图像理解是一个总称。图像识别是一个系统。其中每一部分和其前面的一部分都有一定的关系，也可以说有一种反馈作用，例如分割可以

8、在预处理中进行。并且，该系统不是孤立的，为了发挥其功能，它时时刻刻需要来自外界的必要信息，以便使每个部分能有效地工作。这些外界信息是指处理问题及解决问题的看法、设想、方法等。例如，根据实际图像，在处理部分需要采用什么样的预处理，在识别部分需要怎样分割，抽取什么样的特征及怎样抽取特征，怎样进行分类，要分多少类，以及最后提供结构分析所需的结构信息等。在该系统中，图像预处理、图像分割为图像处理；图像特征提取和图像分类属图像识别；而结构句法分析涉及的内容则是从图像分割到图像结构分析这一过程；整个系统所得到的结果是图像的描述及解释。当某个新的对象（图像）送进系统时，就可以进行解释，说明它是什么。8.Ma

9、tlab技术简介Mat lab是Math works公司与1982年推出的一套高性能的数值计算和可视化数学软件。被誉为“巨人肩上的工具”。由于使用Mat lab编程运算与人进行科学计算的思路和表达方式完全一致，所以不像学习其它高级语言，如Basic和C等那样难于掌握，用Mat lab编写程序犹如在演算纸上排列出公式与求解问题，所以又被称为演算科学算法语言。一般数值分析、矩阵运算、数字信号处理、建模和系统控制和优化等应用程序，并集应用程序和图形于一个便于使用的集成环境中。在这个环境下，对所要求解的问题，用户只需简单地列出数学表达式，其结果便以数值或图形方式显示出来。Mat lab的含义是矩阵实验

10、室(MATRIXLABORATORY)，主要用于方便矩阵的存取，其基本元素是无须定义维数的矩阵。Mat lab自问世以来，就是以数值计算称雄。Mat lab进行数值计算的基本单位是复数数组(或称阵列)，这使得Mat lab高度“向量化”。经过十几年的完善和扩充，它现已发展成为线性代数课程的标准工具。由于它不需定义数组的维数，并给出矩阵函数、特殊矩阵专门的库函数，使之在求解诸如信号处理、建模、系统识别、控制、优化等领域的问题时，显得大为简捷、高效、方便，这是其它高级语言所不能比拟的。美国许多大学的实验室都安装有Mat lab供学习和研究之用。在那里，Mat lab是攻读学位的大学生、硕士生、博士

11、生必须掌握的基本工具。Mat lab中包括了被称作工具箱(TOOU玉OX)的各类应用问题的求解工具。工具箱实际上是对Mat lab进行扩展应用的一系列Mat lab函数(称为M文件)，它可用来求解各类学科的问题，包括信号处理、图象处理、控制系统辨识、神经网络等。随着Mat lab版本的不断升级，其所含的工具箱的功能也越来越丰富，因此，应用范围也越来越广泛，成为涉及数值分析的各类工程师不可不用的工具。Image Science ReviewIn recent years, image processing and recognition technology has developed rapi

12、dly, now people have been fully aware of image processing and recognition technology is to understand the world and an important means of transforming the world. Currently, it has been used in many fields, 21 century information age an important high-tech science and technology.1. Image Processing a

13、nd Recognition TechnologyImage is to use all kinds of observing systems in various forms and means of observation and access to the objective world, directly or indirectly, in the role of a human eye and visual perception of the entity. Scientific research and statistics show that the human race fro

14、m the outside world the information obtained about 75 percent comes from the visual system, that is, human beings are the most of the information obtained from the image.Image processing is an important extension of human vision means that people can be arbitrary wavelength measured by the image. Wi

15、th CT can see objects within the fault image with the corresponding tools can be seen as three-dimensional images and post-image. 1964, the United States in space exploration shoot the moon to a large number of photos, but because of various environmental factors, these photos is not very clear, to

16、that end, the U.S. Jet Propulsion Laboratory (JPL) on the use of computer image processing, Photos of important information to clear reproduction. This is the door an important milestone in the development of technology. Generally speaking, the development of image processing technologies generally

17、experienced a start-up phase, the period of development, popularization and application of view of four stages. Start-up phase began in the 1960s, the image grating used to scan revealed that most of a towel, its mainframe processing. During this period, due to the high cost of image storage, proces

18、sing equipment costs, thus its use of very narrow. In the 1970s entered a period of development, began using large, image processing is gradually switching to raster scan display, especially in the CT and satellite remote sensing image, the image processing technology development played a very good

19、Facilitating role. By the 1980s, image processing technology into the popular view, this purchase has been able to take on computer graphics image processing tasks. VLSL makes the emergence of more processing speed greatly improved, further reducing its cost, greatly promote the popularity of graphi

20、cs imaging systems and applications. 20 images in the 1990s is the period of practical technology, image processing of the huge amount of information on the requirements of high processing speed.21st century technology to high-quality images of the area of development, embodied in the following poin

21、ts: high-resolution, high speed, image processing technology development of the ultimate goal is to achieve real-time image processing, which in the formation of a moving target, identification And on the track is of great significance: three-dimensional, three-dimensional information covered by the

22、 most complete and rich, digital holographic technology will help to achieve this objective; intelligent, and its purpose is to achieve a smart image formation, processing, recognition and understanding .2. Image processing and recognition technology applicationsAt present, image processing and reco

23、gnition technology the main areas of biomedical applications, document processing, industrial inspection, robot vision, cargo inspection, zip code, finance, public security, banking, machinery, transport, e-commerce and multimedia communication networks, and other fields.3. Image processing and reco

24、gnitionDigital image processing and identifying the knowledge disciplines involved in a very wide variety of specific ways, the application also very common, but research on the subject can be divided into the following areas: digital image, image transformation, image enhancement, image segmentatio

25、n.4. Image Recognition TechnologyRecognition is nearly 20 years to develop a new type of science and technology, to study certain object or process (collectively images) and the classification described as the main content. Image Recognition by the study a wide variety of areas, it can be a medical

26、image of the cancer cell recognition; machining parts and components in the identification, classification can be identified in remote sensing images to identify crops, forests, lakes and military facilities, and the judge of crops Growth is forecast harvest levels; can be a self-guided car in the p

27、ath recognition; postal system of automatic letter sorting; traffic control, identification of regulations on vehicle, bank identification, ID cards and other identification. These images recognition of are the subject. Overall, the studys main problem is classification issues.5. Image ProcessingIn

28、examining images, the first to receive the image information on the pretreatment (before processing) filtered to interference, noise, for geometry, color correction, and so on. This will improve the signal to noise ratio; sometimes because of weak information, is not recognized, had to deal with enh

29、ancements. Enhance the role is to provide a meet certain requirements of the image, or images transform, or to computer analysis. And in order to find images from the need to identify things have to split the image, that is, positioning and separation, separation of different objects. To give observ

30、ers a clear image, but also the image of good governance, that is, rehabilitation treatment, it is already degraded the image to rebuild or restore the process in order to improve the image fidelity. In the actual processing, because image is very large amount of information in storage and transmiss

31、ion, but also the image data compression.The work must be carried out by computer, thus to encode and so on. The role is the least number of digital coding (also known as bits), said monochrome and color image for a more effective transmission and storage. The above are considered image processing a

32、reas. Therefore, image processing, including image coding, image enhancement, image compression, image restoration, image segmentation, and so on. On the part of image processing, image processing, the main purpose is to solve two problems: First, determine whether the image needed: another is to de

33、termine what information is.6. Image UnderstandingThe so-called image understanding is a general term. The image processing and image recognition of the ultimate goal is to make the image described and explained that in order to understand what it is the final image. So it is in image processing and

34、 image recognition on the basis of re-classification under the syntactic structure for analysis, and interpretation of images to describe images. Thus image understanding, including image processing, image recognition and analysis.7. The relationship of Image recognition and image processing and ima

35、ge understandingAbove that image is a general term for understanding. Recognition is a system. Each of these parts and the front part of a certain relationship can also say there is a feedback effect, for example, can be divided in the pretreatment of conduct. In addition, it always needs the necess

36、ary information from the outside world, so that each part can work effectively. For example, according to the actual images, some need to adopt in dealing with what kind of preprocessing, in recognition of the need to separate what, from what kind of features and characteristics of how to take and h

37、ow to classify, to the number of sub-category, and, finally, to provide the necessary structural analysis The structure of information.In this system, image pre-processing, image segmentation for image processing, image feature extraction and image classification of image recognition and the syntact

38、ic structure of the content is separated from image to image the structure of this process; system as a whole received The result is the description and image interpretation. When a new object (images) into the system, you can interpret that what it is.8. Mat lab Technical OverviewMat lab is Math wo

39、rks with the 1982 launch of a high-performance numerical computing and visualization of mathematical software. As the use of computing and programming Mat lab conducted scientific computing ideas and expression exactly the same, so unlike other high-level language learning, such as Basic and C, and

40、so on, as difficult to grasp, with Mat lab procedures for the preparation of the calculus is like a piece of paper with formula And solving problems, is also known as calculus scientific.General numerical analysis, matrix computing, digital signal processing, modeling and system control and optimiza

41、tion applications, and collect applications and graphics in a user-friendly integrated environment. In this environment, the solution of the problem required, users simply set out mathematical expression, the result will be numerical or graphic to show.Mat lab is the meaning of Matrix Laboratories (

42、MATRIXLABORATORY), the basic element is not defined dimension of the matrix. Mat lab has since come out. Mat lab numerical calculation is the basic unit complex array (or array), which makes high Mat lab to quantify. After more than 10 years of refinement and expansion, it has become the standard li

43、near algebra course tools. Because no definition of an array of dimension, and gives matrix function, special matrix specialized library function, so that in solving such as signal processing, modeling, system identification, control and optimize the areas of the issue, it Greatly simple, efficient,

44、 convenient and other senior This is unprecedented language. Many American universities have installed a laboratory Mat lab for study and research purposes. There, doctoral students must master the basic tools. Mat lab, known as included in the toolbox application of the various tools to solve the p

45、roblem. Kit is actually Mat lab to expand the application of a series of Mat lab function it can be used to solve various disciplines, including signal processing, image processing, the control system identification, neural networks. With Mat lab version of the continually escalating, the functions contained in the toolbox and extensive Therefore, the application of more and more extensive, involving a numerical analysis of the types of engineers do not have to be a tool.