網(wǎng)絡(luò)工程畢業(yè)論文外文翻譯-數(shù)字圖像處理與邊緣檢測(cè)

1、<p><b>　　班 級(jí) </b></p><p><b>　　學(xué)號(hào) </b></p><p><b>　　本科畢業(yè)設(shè)計(jì)論文</b></p><p>　　題 目： 掃描儀色彩增強(qiáng)及可視化系統(tǒng)設(shè)計(jì)與實(shí)現(xiàn) </p><p>

2、;　　學(xué) 院： 計(jì)算機(jī)學(xué)院 </p><p>　　專(zhuān) 業(yè)： 網(wǎng)絡(luò)工程 </p><p>　　學(xué)生姓名： </p><p>　　導(dǎo)師姓名：

3、 </p><p>　　Digital Image Processing and Edge Detection</p><p>　　Digital Image Processing</p><p>　　Interest in digital image processing methods stems from two principal applic

4、a- tion areas: improvement of pictorial information for human interpretation; and processing of image data for storage, transmission, and representation for au- tonomous machine perception. </p><p>

5、;　　An image may be defined as a two-dimensional function, f(x, y), where x and y are spatial (plane) coordinates, and the amplitude of f at any pair of coordinates (x, y) is called the intensity or gray level of the

6、 image at that point. When x, y, and the amplitude values of f are all finite, discrete quantities, we call the image a digital image. The field of digital image processing refers to processing digital images by means

7、 of a digital computer. Note that a digital image is composed </p><p>　　Vision is the most advanced of our senses, so it is not surprising that images play the single most important role in human perce

8、ption. However, unlike humans, who are limited to the visual band of the electromagnetic (EM) spec- trum, imaging machines cover almost the entire EM spectrum, ranging from gamma to radio waves. They can operat

9、e on images generated by sources that humans are not accustomed to associating with images. These include ultra- sound, electron microscop</p><p>　　There is no general agreement among authors reg

10、arding where image processing stops and other related areas, such as image analysis and computer vi- sion, start. Sometimes a distinction is made by defining image processing as a discipline in which both the inp

11、ut and output of a process are images. We believe this to be a limiting and somewhat artificial boundary. For example, under this definition, even the trivial task of computing the average intensity of an image (

12、which yie</p><p>　　There are no clearcut boundaries in the continuum from image processing at one end to computer vision at the other. However, one useful paradigm is to consider three types of comp

13、uterized processes in this continuum: low-, mid-, and highlevel processes. Low-level processes involve primitive opera- tions such as image preprocessing to reduce noise, contrast enhancement, and image sharpeni

14、ng. A low-level process is characterized by the fact that both its inputs and outputs </p><p>　　Based on the preceding comments, we see that a logical place of overlap between image processing and i

15、mage analysis is the area of recognition of individual regions or objects in an image. Thus, what we call in this book digital image processing encompasses processes whose inputs and outputs are images and, in ad

16、dition, encompasses processes that extract attributes from images, up to and including the recognition of individual objects. As a simple illustration to clarify these c</p><p>　　The areas of applicati

17、on of digital image processing are so varied that some form of organization is desirable in attempting to capture the breadth of this field. One of the simplest ways to develop a basic understanding of the exten

18、t of image processing applications is to categorize images according to their source (e.g., visual, X-ray, and so on). The principal energy source for images in use today is the electromagnetic energy spectrum. Ot

19、her important sources of energy incl</p><p>　　Images based on radiation from the EM spectrum are the most familiar, es- pecially images in the X-ray and visual bands of the spectrum. Electromagnet- ic wa

20、ves can be conceptualized as propagating sinusoidal waves of varying wavelengths, or they can be thought of as a stream of massless particles, each traveling in a wavelike pattern and moving at the speed of lig

21、ht. Each massless particle contains a certain amount (or bundle) of energy. Each bundle of energy is called a photo</p><p>　　Image acquisition is the first process. Note that acquisition could be as

22、simple as being given an image that is already in digital form. Generally, the image acquisition stage involves preprocessing, such as scaling.</p><p>　　Image enhancement is among the simplest and most a

23、ppealing areas of digital image processing. Basically, the idea behind enhancement techniques is to bring out detail that is obscured, or simply to highlight certain features of interest in an image. A familiar exa

24、mple of enhancement is when we increase the contrast of an image because “it looks better.” It is important to keep in mind that enhancement is a very subjective area of image processing. Image restoration is an a

25、rea that</p><p>　　Color image processing is an area that has been gaining in importance because of the significant increase in the use of digital images over the Internet. It covers a number of fundam

26、ental concepts in color models and basic color processing in a digital domain. Color is used also in later chapters as the basis for extracting features of interest in an image.</p><p>　　Wavelets are

27、the foundation for representing images in various degrees of resolution. In particular, this material is used in this book for image data compression and for pyramidal representation, in which images are subdi

28、vided successively into smaller regions.</p><p>　　Compression, as the name implies, deals with techniques for reducing the storage required to save an image, or the bandwidth required to transmi it.Although

29、storage technology has improved significantly over the past decade, the same cannot be said for transmission capacity. This is true particularly in uses of the Internet, which are characterized by significant pict

30、orial content. Image compression is familiar (perhaps inadvertently) to most users of computers in the form of image</p><p>　　Morphological processing deals with tools for extracting image component

31、s that are useful in the representation and description of shape. The material in this chapter begins a transition from processes that output images to processes that output image attributes.</p><p>

32、　　Segmentation procedures partition an image into its constituent parts or objects. In general, autonomous segmentation is one of the most difficult tasks in digital image processing. A rugged segmentation procedu

33、re brings the process a long way toward successful solution of imaging problems that require objects to be identified individually. On the other hand, weak or erratic segmentation algorithms almost always guarante

34、e eventual failure. In general, the more accurate the segme</p><p>　　Representation and description almost always follow the output of a segmentation stage, which usually is raw pixel data, constitut

35、ing either the bound- ary of a region (i.e., the set of pixels separating one image region from another) or all the points in the region itself. In either case, converting the data to a form suitable for computer p

36、rocessing is necessary. The first decision that must be made is whether the data should be represented as a boundary or as a complete r</p><p>　　Recognition is the process that assigns a label (

37、e.g., “vehicle”) to an object based on its descriptors. As detailed before, we conclude our coverage of digital image processing with the development of methods for recognition of individual objects.</p>&

38、lt;p>　　So far we have said nothing about the need for prior knowledge or about the interaction between the knowledge base and the processing modules in Fig2 above. Knowledge about a problem domain is co

39、ded into an image processing system in the form of a knowledge database. This knowledge may be as sim- ple as detailing regions of an image where the information of interest is known to be located, thus limiting the

40、search that has to be conducted in seeking that information. The kno</p><p>　　Edge detection</p><p>　　Edge detection is a terminology in image processing and computer vision, particularly in t

41、he areas of feature detection and feature extraction, to refer to algorithms which aim at identifying points in a digital image at which the image brightness changes sharply or more formally has discontinuities.Although

42、point and line detection certainly are important in any discussion on segmentation,edge dectection is by far the most common approach for detecting meaningful discounties in gray level.</p><p>　　Although cer

43、tain literature has considered the detection of ideal step edges, the edges obtained from natural images are usually not at all ideal step edges. Instead they are normally affected by one or several of the following effe

44、cts:1.focal blur caused by a finite depth-of-field and finite point spread function; 2.penumbral blur caused by shadows created by light sources of non-zero radius; 3.shading at a smooth object edge; 4.local specularitie

45、s or interreflections in the vicinity of object </p><p>　　A typical edge might for instance be the border between a block of red color and a block of yellow. In contrast a line (as can be extracted by a ridg

46、e detector) can be a small number of pixels of a different color on an otherwise unchanging background. For a line, there may therefore usually be one edge on each side of the line.</p><p>　　To illustrate wh

47、y edge detection is not a trivial task, let us consider the problem of detecting edges in the following one-dimensional signal. Here, we may intuitively say that there should be an edge between the 4th and 5th pixels.<

48、;/p><p>　　If the intensity difference were smaller between the 4th and the 5th pixels and if the intensity differences between the adjacent neighbouring pixels were higher, it would not be as easy to say that t

49、here should be an edge in the corresponding region. Moreover, one could argue that this case is one in which there are several edges.Hence, to firmly state a specific threshold on how large the intensity change between t

50、wo neighbouring pixels must be for us to say that there should be an edge betwee</p><p>　　There are many methods for edge detection, but most of them can be grouped into two categories,search-based and zero-

51、crossing based. The search-based methods detect edges by first computing a measure of edge strength, usually a first-order derivative expression such as the gradient magnitude, and then searching for local directional ma

52、xima of the gradient magnitude using a computed estimate of the local orientation of the edge, usually the gradient direction. The zero-crossing based methods sear</p><p>　　The edge detection methods that ha

53、ve been published mainly differ in the types of smoothing filters that are applied and the way the measures of edge strength are computed. As many edge detection methods rely on the computation of image gradients, they a

54、lso differ in the types of filters used for computing gradient estimates in the x- and y-directions.</p><p>　　Once we have computed a measure of edge strength (typically the gradient magnitude), the next sta

55、ge is to apply a threshold, to decide whether edges are present or not at an image point. The lower the threshold, the more edges will be detected, and the result will be increasingly susceptible to noise, and also to pi

56、cking out irrelevant features from the image. Conversely a high threshold may miss subtle edges, or result in fragmented edges.</p><p>　　If the edge thresholding is applied to just the gradient magnitude ima

57、ge, the resulting edges will in general be thick and some type of edge thinning post-processing is necessary. For edges detected with non-maximum suppression however, the edge curves are thin by definition and the edge p

58、ixels can be linked into edge polygon by an edge linking (edge tracking) procedure. On a discrete grid, the non-maximum suppression stage can be implemented by estimating the gradient direction using first-orde</p>

59、<p>　　A commonly used approach to handle the problem of appropriate thresholds for thresholding is by using thresholding with hysteresis. This method uses multiple thresholds to find edges. We begin by using the u

60、pper threshold to find the start of an edge. Once we have a start point, we then trace the path of the edge through the image pixel by pixel, marking an edge whenever we are above the lower threshold. We stop marking our

61、 edge only when the value falls below our lower threshold. This approach </p><p>　　Some edge-detection operators are instead based upon second-order derivatives of the intensity. This essentially captures th

62、e rate of change in the intensity gradient. Thus, in the ideal continuous case, detection of zero-crossings in the second derivative captures local maxima in the gradient.</p><p>　　We can come to a conclusio

63、n that,to be classified as a meaningful edge point,the transition in gray level associated with that point has to be significantly stronger than the background at that point.Since we are dealing with local computations,t

64、he method of choice to determine whether a value is “significant” or not id to use a threshold.Thus we define a point in an image as being as being an edge point if its two-dimensional first-order derivative is greater t

65、han a specified criterion of conne</p><p>　　數(shù)字圖像處理與邊緣檢測(cè)</p><p><b>　　數(shù)字圖像處理</b></p><p>　　數(shù)字圖像處理方法的研究源于兩個(gè)主要應(yīng)用領(lǐng)域：其一是為了便于人們分析而對(duì)圖像信息進(jìn)行改進(jìn)：其二是為使機(jī)器自動(dòng)理解而對(duì)圖像數(shù)據(jù)進(jìn)行存儲(chǔ)、傳輸及顯示。</p>

66、;<p>　　一幅圖像可定義為一個(gè)二維函數(shù)f(x,y)，這里x和y是空間坐標(biāo)，而在任何一對(duì)空間坐標(biāo)（x,y）上的幅值f 稱(chēng)為該點(diǎn)圖像的強(qiáng)度或灰度。當(dāng)x,y和幅值f為有限的、離散的數(shù)值時(shí)，稱(chēng)該圖像為數(shù)字圖像。數(shù)字圖像處理是指借用數(shù)字計(jì)算機(jī)處理數(shù)字圖像，值得提及的是數(shù)字圖像是由有限的元素組成的，每一個(gè)元素都有一個(gè)特定的位置和幅值，這些元素稱(chēng)為圖像元素、畫(huà)面元素或像素。像素是廣泛用于表示數(shù)字圖像元素的詞匯。</p>

67、<p>　　視覺(jué)是人類(lèi)最高級(jí)的感知器官，所以，毫無(wú)疑問(wèn)圖像在人類(lèi)感知中扮演著最重要的角色。然而，人類(lèi)感知只限于電磁波譜的視覺(jué)波段，成像機(jī)器則可覆蓋幾乎全部電磁波譜，從伽馬射線(xiàn)到無(wú)線(xiàn)電波。它們可以對(duì)非人類(lèi)習(xí)慣的那些圖像源進(jìn)行加工，這些圖像源包括超聲波、電子顯微鏡及計(jì)算機(jī)產(chǎn)生的圖像。因此，數(shù)字圖像處理涉及各種各樣的應(yīng)用領(lǐng)域。</p><p>　　圖像處理涉及的范疇或其他相關(guān)領(lǐng)域（例如，圖像分析和計(jì)算機(jī)視覺(jué)

68、）的界定在初創(chuàng)人之間并沒(méi)有一致的看法。有時(shí)用處理的輸入和輸出內(nèi)容都是圖像這一特點(diǎn)來(lái)界定圖像處理的范圍。我們認(rèn)為這一定義僅是人為界定和限制。例如，在這個(gè)定義下，甚至最普通的計(jì)算一幅圖像灰度平均值的工作都不能算做是圖像處理。另一方面，有些領(lǐng)域（如計(jì)算機(jī)視覺(jué)）研究的最高目標(biāo)是用計(jì)算機(jī)去模擬人類(lèi)視覺(jué)，包括理解和推理并根據(jù)視覺(jué)輸入采取行動(dòng)等。這一領(lǐng)域本身是人工智能的分支，其目的是模仿人類(lèi)智能。人工智能領(lǐng)域處在其發(fā)展過(guò)程中的初期階段，它的發(fā)展比預(yù)期

69、的要慢的多，圖像分析（也稱(chēng)為圖像理解）領(lǐng)域則處在圖像處理和計(jì)算機(jī)視覺(jué)兩個(gè)學(xué)科之間。</p><p>　　從圖像處理到計(jì)算機(jī)視覺(jué)這個(gè)連續(xù)的統(tǒng)一體內(nèi)并沒(méi)有明確的界線(xiàn)。然而，在這個(gè)連續(xù)的統(tǒng)一體中可以考慮三種典型的計(jì)算處理（即低級(jí)、中級(jí)和高級(jí)處理）來(lái)區(qū)分其中的各個(gè)學(xué)科。</p><p>　　低級(jí)處理涉及初級(jí)操作，如降低噪聲的圖像預(yù)處理，對(duì)比度增強(qiáng)和圖像尖銳化。低級(jí)處理是以輸入、輸出都是圖像為特點(diǎn)的

70、處理。中級(jí)處理涉及分割（把圖像分為不同區(qū)域或目標(biāo)物）以及縮減對(duì)目標(biāo)物的描述，以使其更適合計(jì)算機(jī)處理及對(duì)不同目標(biāo)的分類(lèi)（識(shí)別）。中級(jí)圖像處理是以輸入為圖像，但輸出是從這些圖像中提取的特征（如邊緣、輪廓及不同物體的標(biāo)識(shí)等）為特點(diǎn)的。最后，高級(jí)處理涉及在圖像分析中被識(shí)別物體的總體理解，以及執(zhí)行與視覺(jué)相關(guān)的識(shí)別函數(shù)（處在連續(xù)統(tǒng)一體邊緣）等。</p><p>　　根據(jù)上述討論，我們看到，圖像處理和圖像分析兩個(gè)領(lǐng)域合乎邏輯的

71、重疊區(qū)域是圖像中特定區(qū)域或物體的識(shí)別這一領(lǐng)域。這樣，在研究中，我們界定數(shù)字圖像處理包括輸入和輸出均是圖像的處理，同時(shí)也包括從圖像中提取特征及識(shí)別特定物體的處理。舉一個(gè)簡(jiǎn)單的文本自動(dòng)分析方面的例子來(lái)具體說(shuō)明這一概念。在自動(dòng)分析文本時(shí)首先獲取一幅包含文本的圖像，對(duì)該圖像進(jìn)行預(yù)處理，提?。ǚ指睿┳址?，然后以適合計(jì)算機(jī)處理的形式描述這些字符，最后識(shí)別這些字符，而所有這些操作都在本文界定的數(shù)字圖像處理的范圍內(nèi)。理解一頁(yè)的內(nèi)容可能要根據(jù)理解的復(fù)雜度

72、從圖像分析或計(jì)算機(jī)視覺(jué)領(lǐng)域考慮問(wèn)題。這樣，我們定義的數(shù)字圖像處理的概念將在有特殊社會(huì)和經(jīng)濟(jì)價(jià)值的領(lǐng)域內(nèi)通用。</p><p>　　數(shù)字圖像處理的應(yīng)用領(lǐng)域多種多樣，所以文本在內(nèi)容組織上盡量達(dá)到該技術(shù)應(yīng)用領(lǐng)域的廣度。闡述數(shù)字圖像處理應(yīng)用范圍最簡(jiǎn)單的一種方法是根據(jù)信息源來(lái)分類(lèi)（如可見(jiàn)光、X射線(xiàn)，等等）。在今天的應(yīng)用中，最主要的圖像源是電磁能譜，其他主要的能源包括聲波、超聲波和電子（以用于電子顯微鏡方法的電子束形式）。建

73、模和可視化應(yīng)用中的合成圖像由計(jì)算機(jī)產(chǎn)生。</p><p>　　建立在電磁波譜輻射基礎(chǔ)上的圖像是最熟悉的，特別是X射線(xiàn)和可見(jiàn)光譜圖像。電磁波可定義為以各種波長(zhǎng)傳播的正弦波，或者認(rèn)為是一種粒子流，每個(gè)粒子包含一定（一束）能量，每束能量成為一個(gè)光子。如果光譜波段根據(jù)光譜能量進(jìn)行分組，我們會(huì)得到下圖1所示的伽馬射線(xiàn)（最高能量）到無(wú)線(xiàn)電波（最低能量）的光譜。如圖所示的加底紋的條帶表達(dá)了這樣一個(gè)事實(shí)，即電磁波譜的各波段間并沒(méi)

74、有明確的界線(xiàn)，而是由一個(gè)波段平滑地過(guò)渡到另一個(gè)波段。</p><p>　　圖像獲取是第一步處理。注意到獲取與給出一幅數(shù)字形式的圖像一樣簡(jiǎn)單。通常，圖像獲取包括如設(shè)置比例尺等預(yù)處理。</p><p>　　圖像增強(qiáng)是數(shù)字圖像處理最簡(jiǎn)單和最有吸引力的領(lǐng)域?；旧?，增強(qiáng)技術(shù)后面的思路是顯現(xiàn)那些被模糊了的細(xì)節(jié)，或簡(jiǎn)單地突出一幅圖像中感興趣的特征。一個(gè)圖像增強(qiáng)的例子是增強(qiáng)圖像的對(duì)比度，使其看起來(lái)好一些

75、。應(yīng)記住，增強(qiáng)是圖像處理中非常主觀的領(lǐng)域，這一點(diǎn)很重要。</p><p>　　圖像復(fù)原也是改進(jìn)圖像外貌的一個(gè)處理領(lǐng)域。然而，不像增強(qiáng)，圖像增強(qiáng)是主觀的，而圖像復(fù)原是客觀的。在某種意義上說(shuō)，復(fù)原技術(shù)傾向于以圖像退化的數(shù)學(xué)或概率模型為基礎(chǔ)。另一方面，增強(qiáng)以怎樣構(gòu)成好的增強(qiáng)效果這種人的主觀偏愛(ài)為基礎(chǔ)。</p><p>　　彩色圖像處理已經(jīng)成為一個(gè)重要領(lǐng)域，因?yàn)榛诨ヂ?lián)網(wǎng)的圖像處理應(yīng)用在不斷增長(zhǎng)。

76、就使得在彩色模型、數(shù)字域的彩色處理方面涵蓋了大量基本概念。在后續(xù)發(fā)展，彩色還是圖像中感興趣特征被提取的基礎(chǔ)。</p><p>　　小波是在各種分辨率下描述圖像的基礎(chǔ)。特別是在應(yīng)用中，這些理論被用于圖像數(shù)據(jù)壓縮及金字塔描述方法。在這里，圖像被成功地細(xì)分為較小的區(qū)域。</p><p>　　壓縮，正如其名稱(chēng)所指的意思，所涉及的技術(shù)是減少圖像的存儲(chǔ)量，或者在傳輸圖像時(shí)降低頻帶。雖然存儲(chǔ)技術(shù)在過(guò)去的

77、十年內(nèi)有了很大改進(jìn)，但對(duì)傳輸能力我們還不能這樣說(shuō)，尤其在互聯(lián)網(wǎng)上更是如此，互聯(lián)網(wǎng)是以大量的圖片內(nèi)容為特征的。圖像壓縮技術(shù)對(duì)應(yīng)的圖像文件擴(kuò)展名對(duì)大多數(shù)計(jì)算機(jī)用戶(hù)是很熟悉的（也許沒(méi)注意），如JPG文件擴(kuò)展名用于JPEG（聯(lián)合圖片專(zhuān)家組）圖像壓縮標(biāo)準(zhǔn)。</p><p>　　形態(tài)學(xué)處理設(shè)計(jì)提取圖像元素的工具，它在表現(xiàn)和描述形狀方面非常有用。這一章的材料將從輸出圖像處理到輸出圖像特征處理的轉(zhuǎn)換開(kāi)始。</p>

78、<p>　　分割過(guò)程將一幅圖像劃分為組成部分或目標(biāo)物。通常，自主分割是數(shù)字圖像處理中最為困難的任務(wù)之一。復(fù)雜的分割過(guò)程導(dǎo)致成功解決要求物體被分別識(shí)別出來(lái)的成像問(wèn)題需要大量處理工作。另一方面，不健壯且不穩(wěn)定的分割算法幾乎總是會(huì)導(dǎo)致最終失敗。通常，分割越準(zhǔn)確，識(shí)別越成功。</p><p>　　表示和描述幾乎總是跟隨在分割步驟的輸后邊，通常這一輸出是未加工的數(shù)據(jù)，其構(gòu)成不是區(qū)域的邊緣（區(qū)分一個(gè)圖像區(qū)域和另一個(gè)

79、區(qū)域的像素集）就是其區(qū)域本身的所有點(diǎn)。無(wú)論哪種情況，把數(shù)據(jù)轉(zhuǎn)換成適合計(jì)算機(jī)處理的形式都是必要的。首先，必須確定數(shù)據(jù)是應(yīng)該被表現(xiàn)為邊界還是整個(gè)區(qū)域。當(dāng)注意的焦點(diǎn)是外部形狀特性（如拐角和曲線(xiàn)）時(shí)，則邊界表示是合適的。當(dāng)注意的焦點(diǎn)是內(nèi)部特性（如紋理或骨骼形狀）時(shí)，則區(qū)域表示是合適的。則某些應(yīng)用中，這些表示方法是互補(bǔ)的。選擇一種表現(xiàn)方式僅是解決把原始數(shù)據(jù)轉(zhuǎn)換為適合計(jì)算機(jī)后續(xù)處理的形式的一部分。為了描述數(shù)據(jù)以使感興趣的特征更明顯，還必須確定一種方

80、法。描述也叫特征選擇，涉及提取特征，該特征是某些感興趣的定量信息或是區(qū)分一組目標(biāo)與其他目標(biāo)的基礎(chǔ)。</p><p>　　識(shí)別是基于目標(biāo)的描述給目標(biāo)賦以符號(hào)的過(guò)程。如上文詳細(xì)討論的那樣，我們用識(shí)別個(gè)別目標(biāo)方法的開(kāi)發(fā)推出數(shù)字圖像處理的覆蓋范圍。</p><p>　　到目前為止，還沒(méi)有談到上面圖2中關(guān)于先驗(yàn)知識(shí)及知識(shí)庫(kù)與處理模塊之間的交互這部分內(nèi)容。關(guān)于問(wèn)題域的知識(shí)以知識(shí)庫(kù)的形式被編碼裝入一個(gè)圖

81、像處理系統(tǒng)。這一知識(shí)可能如圖像細(xì)節(jié)區(qū)域那樣簡(jiǎn)單，在這里，感興趣的信息被定位，這樣，限制性的搜索就被引導(dǎo)到尋找的信息處。知識(shí)庫(kù)也可能相當(dāng)復(fù)雜，如材料檢測(cè)問(wèn)題中所有主要缺陷的相關(guān)列表或者圖像數(shù)據(jù)庫(kù)（該庫(kù)包含變化檢測(cè)應(yīng)用相關(guān)區(qū)域的高分辨率衛(wèi)星圖像）。除了引導(dǎo)每一個(gè)處理模塊的操作，知識(shí)庫(kù)還要控制模塊間的交互。這一特性上面圖2中的處理模塊和知識(shí)庫(kù)間用雙箭頭表示。相反單頭箭頭連接處理模塊。</p><p><b>

82、　　邊緣檢測(cè)</b></p><p>　　邊緣檢測(cè)是圖像處理和計(jì)算機(jī)視覺(jué)中的術(shù)語(yǔ)，尤其在特征檢測(cè)和特征抽取領(lǐng)域，是一種用來(lái)識(shí)別數(shù)字圖像亮度驟變點(diǎn)即不連續(xù)點(diǎn)的算法。盡管在任何關(guān)于分割的討論中，點(diǎn)和線(xiàn)檢測(cè)都是很重要的，但是邊緣檢測(cè)對(duì)于灰度級(jí)間斷的檢測(cè)是最為普遍的檢測(cè)方法。</p><p>　　雖然某些文獻(xiàn)提過(guò)理想的邊緣檢測(cè)步驟，但自然界圖像的邊緣并不總是理想的階梯邊緣。相反，它們通

83、常受到一個(gè)或多個(gè)下面所列因素的影響：1.有限場(chǎng)景深度帶來(lái)的聚焦模糊；2.非零半徑光源產(chǎn)生的陰影帶來(lái)的半影模糊；3.光滑物體邊緣的陰影；4.物體邊緣附近的局部鏡面反射或者漫反射。</p><p>　　一個(gè)典型的邊界可能是（例如）一塊紅色和一塊黃色之間的邊界；與之相反的是邊線(xiàn)，可能是在另外一種不變的背景上的少數(shù)不同顏色的點(diǎn)。在邊線(xiàn)的每一邊都有一個(gè)邊緣。</p><p>　　在對(duì)數(shù)字圖像的處理中

84、，邊緣檢測(cè)是一項(xiàng)非常重要的工作。如果將邊緣認(rèn)為是一定數(shù)量點(diǎn)亮度發(fā)生變化的地方，那么邊緣檢測(cè)大體上就是計(jì)算這個(gè)亮度變化的導(dǎo)數(shù)。為簡(jiǎn)化起見(jiàn)，我們可以先在一維空間分析邊緣檢測(cè)。在這個(gè)例子中，我們的數(shù)據(jù)是一行不同點(diǎn)亮度的數(shù)據(jù)。例如，在下面的1維數(shù)據(jù)中我們可以直觀地說(shuō)在第4與第5個(gè)點(diǎn)之間有一個(gè)邊界：</p><p>　　如果光強(qiáng)度差別比第四個(gè)和第五個(gè)點(diǎn)之間小，或者說(shuō)相鄰的像素點(diǎn)之間光強(qiáng)度差更高，就不能簡(jiǎn)單地說(shuō)相應(yīng)區(qū)域存在邊

85、緣。而且，甚至可以認(rèn)為這個(gè)例子中存在多個(gè)邊緣。除非場(chǎng)景中的物體非常簡(jiǎn)單并且照明條件得到了很好的控制，否則確定一個(gè)用來(lái)判斷兩個(gè)相鄰點(diǎn)之間有多大的亮度變化才算是有邊界的閾值，并不是一件容易的事。實(shí)際上，這也是為什么邊緣檢測(cè)不是一個(gè)簡(jiǎn)單問(wèn)題的原因之一。</p><p>　　有許多用于邊緣檢測(cè)的方法，它們大致可分為兩類(lèi)：基于搜索和基于零交叉.基于搜索的邊緣檢測(cè)方法首先計(jì)算邊緣強(qiáng)度，通常用一階導(dǎo)數(shù)表示，例如梯度模；然后，用

86、計(jì)算估計(jì)邊緣的局部方向，通常采用梯度的方向，并利用此方向找到局部梯度模的最大值?；诹憬徊娴姆椒ㄕ业接蓤D像得到的二階導(dǎo)數(shù)的零交叉點(diǎn)來(lái)定位邊緣。通常用拉普拉斯算子或非線(xiàn)性微分方程的零交叉點(diǎn)，我們將在后面的小節(jié)中描述.濾波做為邊緣檢測(cè)的預(yù)處理通常是必要的，通常采用高斯濾波。</p><p>　　已發(fā)表的邊緣檢測(cè)方法應(yīng)用計(jì)算邊界強(qiáng)度的度量， 這與平滑濾波有本質(zhì)的不同. 正如許多邊緣檢測(cè)方法依賴(lài)于圖像梯度的計(jì)算， 他們用

87、不同種類(lèi)的濾波器來(lái)估計(jì)x-方向和y-方向的梯度.</p><p>　　一旦我們計(jì)算出導(dǎo)數(shù)之后，下一步要做的就是給出一個(gè)閾值來(lái)確定哪里是邊緣位置。閾值越低，能夠檢測(cè)出的邊線(xiàn)越多，結(jié)果也就越容易受到圖片噪聲的影響，并且越容易從圖像中挑出不相關(guān)的特性。與此相反，一個(gè)高的閾值將會(huì)遺失細(xì)的或者短的線(xiàn)段。</p><p>　　如果邊緣閾值應(yīng)用于正確的的梯度幅度圖像，生成的邊緣一般會(huì)較厚，某些形式的邊緣

88、變薄處理是必要的。然而非最大抑制的邊緣檢測(cè)，邊緣曲線(xiàn)的定義十分模糊，邊緣像素可能成為邊緣多邊形通過(guò)一個(gè)邊緣連接（邊緣跟蹤）的過(guò)程。在一個(gè)離散矩陣中，非最大抑制階梯能夠通過(guò)一種方法來(lái)實(shí)現(xiàn)，首先預(yù)測(cè)一階導(dǎo)數(shù)方向、然后把它近似到45度的倍數(shù)、最后在預(yù)測(cè)的梯度方向比較梯度幅度。</p><p>　　一個(gè)常用的這種方法是帶有滯后作用的閾值選擇。這個(gè)方法使用不同的閾值去尋找邊緣。首先使用一個(gè)閾值上限去尋找邊線(xiàn)開(kāi)始的地方。一旦

89、找到了一個(gè)開(kāi)始點(diǎn)，我們?cè)趫D像上逐點(diǎn)跟蹤邊緣路徑，當(dāng)大于門(mén)檻下限時(shí)一直紀(jì)錄邊緣位置，直到數(shù)值小于下限之后才停止紀(jì)錄。這種方法假設(shè)邊緣是連續(xù)的界線(xiàn)，并且我們能夠跟蹤前面所看到的邊緣的模糊部分，而不會(huì)將圖像中的噪聲點(diǎn)標(biāo)記為邊緣。但是，我們?nèi)匀淮嬖谶x擇適當(dāng)?shù)拈撝祬?shù)的問(wèn)題，而且不同圖像的閾值差別也很大。</p><p>　　其它一些邊緣檢測(cè)操作是基于亮度的二階導(dǎo)數(shù)。這實(shí)質(zhì)上是亮度梯度的變化率。在理想的連續(xù)變化情況下，在二

90、階導(dǎo)數(shù)中檢測(cè)過(guò)零點(diǎn)將得到梯度中的局部最大值。另一方面，二階導(dǎo)數(shù)中的峰值檢測(cè)是邊線(xiàn)檢測(cè)，只要圖像操作使用一個(gè)合適的尺度表示。如上所述，邊線(xiàn)是雙重邊緣，這樣我們就可以在邊線(xiàn)的一邊看到一個(gè)亮度梯度，而在另一邊看到相反的梯度。這樣如果圖像中有邊線(xiàn)出現(xiàn)的話(huà)我們就能在亮度梯度上看到非常大的變化。為了找到這些邊線(xiàn)，我們可以在圖像亮度梯度的二階導(dǎo)數(shù)中尋找過(guò)零點(diǎn)。</p><p>　　總之，為了對(duì)有意義的邊緣點(diǎn)進(jìn)行分類(lèi)，與這個(gè)點(diǎn)相

91、聯(lián)系的灰度級(jí)變換必須比在這一點(diǎn)的背景上變換更為有效。由于我們用局部計(jì)算進(jìn)行處理，決定一個(gè)值是否有效的選擇方法就是使用門(mén)限。因此，如果一個(gè)點(diǎn)的二維一階導(dǎo)數(shù)比指定的門(mén)限大，我們就定義圖像中的此點(diǎn)是一個(gè)邊緣點(diǎn)。術(shù)語(yǔ)“邊緣線(xiàn)段”一般在邊緣與圖像的尺寸比起來(lái)很短時(shí)才使用。分割的關(guān)鍵問(wèn)題是如何將邊緣線(xiàn)段組合成更長(zhǎng)的邊緣。如果我們選擇使用二階導(dǎo)數(shù)，則另一個(gè)可用的定義是將圖像中的邊緣點(diǎn)定義為它的二階導(dǎo)數(shù)的零交叉點(diǎn)。此時(shí)，邊緣的定義同上面講過(guò)的定義是一樣

92、的。應(yīng)注意，這些定義并不能保證在一幅圖像中成功地找到邊緣，它們只是給了我們一個(gè)尋找邊緣的形式體系。圖像中的一階導(dǎo)數(shù)用梯度計(jì)算，二階導(dǎo)數(shù)使用拉普拉斯算子得到。</p><p><b>　　指導(dǎo)老師：吳自力</b></p><p><b>　　03072016</b></p><p><b>　　馬維華</b&

93、gt;</p><p>　　e馬,陪F?lm嫄W壙獑??N€?)?0?q??3拞屓??p勌m:b5a漬璺1疆lv~0瘩艌v€??`@2屘b瘌Tq 〆妋4U~K袢,疀薤`,A6玭$? $徤?|ッ?罪=T?G駁op1尚懇袹戦疈u圻?賡O?a?*{p懡0L)w懱拽?E???/扢]?9镃/潺hRＱ烞b蔔Ye嚁劑Bb?捚?慳曥T?UFUZ!醋M[M蘇g壺鰷伸???qfg%厧1&

94、#39;t?€ 戯憤焿驆罳騙^萡P?躀cP縮蒕\?nP砫 }?C?Z豸值?<箾殶竛吐K?M轂?藑琶q@v栃?? 譇>橒忋喏曚N(xiāo)I耔nENB峢mT鸏s眑+?該Ζ|紖?>!汩?膾f抙氄宨澻S箔XM竴玐訨?V?腡?,B衹絵2n吞輷匽e馬e馬,陪F?lm嫄W壙獑??N€?)?0?q??3拞屓??p勌m:b5a漬璺1疆lv~0瘩艌v€??`@2屘b瘌Tq 〆妋4U~K袢,疀薤`