圖像處理外文翻譯

1、<p><b>　　附錄A</b></p><p>　　3 Image Enhancement in the Spatial Domain</p><p>　　The principal objective of enhancement is to process an image so that the result is more suitable th

2、an the original image for a specific application. The word specific is important, because it establishes at the outset than the techniques discussed in this chapter are very much problem oriented. Thus, for example, a me

3、thod that is quite useful for enhancing X-ray images may not necessarily be the best approach for enhancing pictures of Mars transmitted by a space probe. Regardless of the method </p><p>　　Image enhancement

4、 approaches fall into two broad categories: spatial domain methods and frequency domain methods. The term spatial domain refers to the image plane itself, and approaches in this category are based on direct manipulation

5、of pixels in an image. Fourier transform of an image. Spatial methods are covered in this chapter, and frequency domain enhancement is discussed in Chapter 4.Enhancement techniques based on various combinations of method

6、s from these two categories are not unusual</p><p>　　There is no general theory of image enhancement. When an image is processed for visual interpretation, the viewer is the ultimate judge of how well a part

7、icular method works. Visual evaluation of image quality is a highly is highly subjective process, thus making the definition of a “good image” an elusive standard by which to compare algorithm performance. When the probl

8、em is one of processing images for machine perception, the evaluation task is somewhat easier. For example, in dealing with a </p><p>　　3.1 Background</p><p>　　As indicated previously, the term

9、spatial domain refers to the aggregate of pixels composing an image. Spatial domain methods are procedures that operate directly on these pixels. Spatial domain processes will be denotes by the expression</p><

10、p><b>　　(3.1-1)</b></p><p>　　where f(x, y) is the input image, g(x, y) is the processed image, and T is an operator on f, defined over some neighborhood of (x, y). In addition, T can operate on

11、 a set of input images, such as performing the pixel-by-pixel sum of K images for noise reduction, as discussed in Section 3.4.2.</p><p>　　The principal approach in defining a neighborhood about a point (x,

12、y) is to use a square or rectangular subimage area centered at (x, y).The center of the subimage is moved from pixel to starting, say, at the top left corner. The operator T is applied at each location (x, y) to yield th

13、e output, g, at that location. The process utilizes only the pixels in the area of the image spanned by the neighborhood. Although other neighborhood shapes, such as approximations to a circle, sometimes are used</p&g

14、t;<p>　　The simplest from of T is when the neighborhood is of size 1×1 (that is, a single pixel). In this case, g depends only on the value of f at (x, y), and T becomes a gray-level (also called an intensity

15、 or mapping) transformation function of the form</p><p><b>　　(3.1-2)</b></p><p>　　where, for simplicity in notation, r and s are variables denoting, respectively, the grey level of f

16、(x, y) and g(x, y)at any point (x, y).Some fairly simple, yet powerful, processing approaches can be formulates with gray-level transformations. Because enhancement at any point in an image depends only on the grey level

17、 at that point, techniques in this category often are referred to as point processing.</p><p>　　Larger neighborhoods allow considerably more flexibility. The general approach is to use a function of the valu

18、es of f in a predefined neighborhood of (x, y) to determine the value of g at (x, y). One of the principal approaches in this formulation is based on the use of so-called masks (also referred to as filters, kernels, temp

19、lates, or windows). Basically, a mask is a small (say, 3×3) 2-Darray, in which the values of the mask coefficients determine the nature of the type of approach often are</p><p>　　3.2 Some Basic Gray Lev

20、el Transformations</p><p>　　We begin the study of image enhancement techniques by discussing gray-level transformation functions. These are among the simplest of all image enhancement techniques. The values

21、of pixels, before and after processing, will be denoted by r and s, respectively. As indicated in the previous section, these values are related by an expression of the from s = T(r), where T is a transformation that map

22、s a pixel value r into a pixel value s. Since we are dealing with digital quantities, values of the t</p><p>　　As an introduction to gray-level transformations, which shows three basic types of functions use

23、d frequently for image enhancement: linear (negative and identity transformations), logarithmic (log and inverse-log transformations), and power-law (nth power and nth root transformations). The identity function is the

24、trivial case in which out put intensities are identical to input intensities. It is included in the graph only for completeness.</p><p>　　3.2.1 Image Negatives</p><p>　　The negative of an image

25、with gray levels in the range [0, L-1]is obtained by using the negative transformation show shown, which is given by the expression</p><p><b>　　(3.2-1)</b></p><p>　　Reversing the int

26、ensity levels of an image in this manner produces the equivalent of a photographic negative. This type of processing is particularly suited for enhancing white or grey detail embedded in dark regions of an image, especia

27、lly when the black areas are dominant in size. </p><p>　　3.2.2 Log Transformations</p><p>　　The general from of the log transformation is</p><p><b>　　(3.2-2)</b></p&g

28、t;<p>　　Where c is a constant, and it is assumed that r ≥0 .The shape of the log curve transformation maps a narrow range of low gray-level values in the input image into a wider range of output levels. The opposi

29、te is true of higher values of input levels. We would use a transformation of this type to expand the values of dark pixels in an image while compressing the higher-level values. The opposite is true of the inverse log t

30、ransformation.</p><p>　　Any curve having the general shape of the log functions would accomplish this spreading/compressing of gray levels in an image. In fact, the power-law transformations discussed in the

31、 next section are much more versatile for this purpose than the log transformation. However, the log function has the important characteristic that it compresses the dynamic range of image characteristics of spectra. It

32、is not unusual to encounter spectrum values that range from 0 to 106 or higher. While processing </p><p>　　3.2.3 Power-Law Transformations</p><p>　　Power-Law transformations have the basic from&

33、lt;/p><p><b>　　(3.2-3)</b></p><p>　　Where c and y are positive constants .Sometimes Eq. (3.2-3) is written as to account for an offset (that is, a measurable output when the input is

34、zero). However, offsets typically are an issue of display calibration and as a result they are normally ignored in Eq. (3.2-3). Plots of s versus r for various values of y are shown in Fig.3.6. As in the case of the log

35、transformation, power-law curves with fractional values of y map a narrow range of dark input values into a wider range of output v</p><p>　　A variety of devices used for image capture, printing, and display

36、 respond according to as gamma[hence our use of this symbol in Eq.(3.2-3)].The process used to correct this power-law response phenomena is called gamma correction.</p><p>　　Gamma correction is important if

37、displaying an image accurately on a computer screen is of concern. Images that are not corrected properly can look either bleached out, or, what is more likely, too dark. Trying to reproduce colors accurately also requir

38、es some knowledge of gamma correction because varying the value of gamma correcting changes not only the brightness, but also the ratios of red to green to blue. Gamma correction has become increasingly important in the

39、past few years, as use of d</p><p>　　3.2.4 Piecewise-Linear Transformation Functions</p><p>　　A complementary approach to the methods discussed in the previous three sections is to use piecewise

40、 linear functions. The principal advantage of piecewise linear functions over the types of functions we have discussed thus far is that the form of piecewise functions can be arbitrarily complex. In fact, as we will see

41、shortly, a practical implementation of some important transformations can be formulated only as piecewise functions. The principal disadvantage of piecewise functions is that their</p><p>　　Contrast stretchi

42、ng</p><p>　　One of the simplest piecewise linear functions is a contrast-stretching transformation. Low-contrast images can result from poor illumination, lack of dynamic range in the imaging sensor, or even

43、 wrong setting of a lens aperture during image acquisition. The idea behind contrast stretching is to increase the dynamic range of the gray levels in the image being processed.</p><p>　　Gray-level slicing&l

44、t;/p><p>　　Highlighting a specific range of gray levels in an image often is desired. Applications include enhancing features such as masses of water in satellite imagery and enhancing flaws in X-ray images. Th

45、ere are several ways of doing level slicing, but most of them are variations of two basic themes. One approach is to display a high value for all gray levels in the range of interest and a low value for all other gray le

46、vels. </p><p>　　Bit-plane slicing</p><p>　　Instead of highlighting gray-level ranges, highlighting the contribution made to total image appearance by specific bits might be desired. Suppose that

47、 each pixel in an image is represented by 8 bits. Imagine that the image is composed of eight 1-bit planes, ranging from bit-plane 0 for the least significant bit to bit-plane 7 for the most significant bit. In terms of

48、8-bit bytes, plane 0 contains all the lowest order bits in the bytes comprising the pixels in the image and plane 7 contains all </p><p>　　3.3 Histogram Processing</p><p>　　The histogram of a di

49、gital image with gray levels in the range [0, L-1] is a discrete function , where is the kth gray level and is the number of pixels in the image having gray level . It is common practice to pixels in the image, den

50、oted by n. Thus, a normalized histogram is given by , for , Loosely speaking, gives an estimate of the probability of occurrence of gray level . Note that the sum of all components of a normalized histogram is equal

51、 to 1.</p><p>　　Histograms are the basis for numerous spatial domain processing techniques. Histogram manipulation can be used effectively for image enhancement, as shown in this section. In addition to prov

52、iding useful image statistics, we shall see in subsequent chapters that the information inherent in histograms also is quite useful in other image processing applications, such as image compression and segmentation. Hist

53、ograms are simple to calculate in software and also lend themselves to economic hardware </p><p><b>　　附錄B</b></p><p>　　第三章 空間域圖像增強</p><p>　　增強的首要目標(biāo)是處理圖像，使其比原始圖像格式和特定應(yīng)用。&

54、lt;/p><p>　　這里的“特定”很重要，因為它一開始就確立了本章多討論技術(shù)是面向問題的。</p><p>　　例如，一種很合適增強X射線圖像的方法，不一定是增強有空間探測器發(fā)回的火星圖像的最好方法。暫且不談所用方法，圖像增強本身就是圖像處理中最具有吸引力的領(lǐng)域之一。</p><p>　　圖像增強的方法分為兩大類：空間域方法和頻域方法。“空間域”一次是指圖像平面自身

55、，這類方法是以對圖像的像素直接處理恩基礎(chǔ)的?！邦l域”處理技術(shù)足以修改圖像的傅氏變換為基礎(chǔ)的?？臻g域方法在這一章講述，頻域增強將在第四章討論。以這兩類方法的各種結(jié)合為基礎(chǔ)的增強技術(shù)是不常見的。我們也注意到，本章關(guān)于增強的許多基本技術(shù)在后續(xù)章節(jié)里的其他圖像處理應(yīng)用中也會用到。</p><p>　　圖像增強的通用理論是不存在的。當(dāng)圖像為視覺解釋而進行處理時，有觀察者最后判斷特定方法的效果。圖像質(zhì)量的視覺評價是一種高度主

56、觀的過程，因此，定義一個“理想圖像”標(biāo)準(zhǔn)沒通過這個標(biāo)準(zhǔn)去比較算法的性能。當(dāng)為機器感知而處理圖像時，這個評價任務(wù)就會容易一些。例如，在一個特征識別的應(yīng)用中，不考慮像計算要求這些問題，最好的圖像處理方法是一種能得到最好的機器可識別結(jié)果的方法。無論怎樣，甚至在把一個明確的性能標(biāo)準(zhǔn)加于這個問題的情況下，在選擇特定的圖像增強方法之前，常常需要一個實驗和誤差的特定量。</p><p><b>　　3.1背景知識&l

57、t;/b></p><p>　　如前所述，“空間域增強”是指增強構(gòu)成圖像的像素?？臻g域方法是直接對這些像素操作的過程?？臻g域處理可由下式定義：</p><p><b>　　(3.1-1)</b></p><p>　　其中f(x, y)是輸入圖像，g(x, y)是處理后的圖像，T是對f的一種操作，其定義在(x, y)的鄰域。另外，T能對輸入

58、圖像集進行操作，例如，為減少噪音而對K幅圖像進行逐像素的求和操作，如3.4.2節(jié)所討論的。</p><p>　　定義一個點(x, y)鄰域的主要方法是利用中心在(x, y)點的正方形貨矩形子圖像。子圖像的中心從一個像素向另一個像素移動，比如說，可以從左上角開始。T操作應(yīng)用到每一個(x, y)位置得到該店的輸出g。這個過程僅僅用在小范圍鄰域里的圖像像素。盡管像近似于圓的其他鄰域形狀有時也用，但正方形和矩形列陣因其容

59、易執(zhí)行操作而占主導(dǎo)地位。</p><p>　　T操作最簡單的形式是鄰域為1×1的尺度(即單個像素)。在這種情況下，g僅僅依賴于f在(x, y)點的值，T操作成為灰度級變換函數(shù)(也叫做強度映射)，形成為：</p><p><b>　　(3.1-2)</b></p><p>　　這里，為簡便起見，令r和s是所定義的變量，分別是f(x, y

60、)和g(x, y)在任一點(x, y)的灰度級。有的相當(dāng)簡單，卻有很大作用，處理方法可以用灰度變換加以公式化。因為在圖像任意點的增強僅僅依賴于該點的灰度，這類技術(shù)常常是指點處理。</p><p>　　更大的鄰域會有更多的靈活性。一般的方法是，利用點(x, y)事先定義的鄰域里的一個f值的函數(shù)來決定g在(x, y)的值，其公式化的一個主要方法是以利用所謂的模板(也指濾波器、核、掩模或窗口)為基礎(chǔ)的。從根本上說，模板

61、是一個小的(即3×3)二維陣列，模板的系數(shù)值決定了處理的性質(zhì)，如圖像尖銳化等。以這種方法為基礎(chǔ)的增強技術(shù)通常是指模板處理或濾波。這些概念將在3.5節(jié)討論。</p><p>　　3.2某些基本灰度變換</p><p>　　以討論灰度變換函數(shù)開始研究圖像增強技術(shù)，這些都屬于所有圖像增強技術(shù)最簡單的一類。處理前后的像素的值用r和s分別定義。如前節(jié)所述，這些值與s = T(r)表達式的形

62、式有關(guān)，這里的T是把像素的值r映射到值s的一種變換。由于處理的是數(shù)字量，變換函數(shù)的值通常儲存在一個一維陣列中，并且從r到s的映射通過查表得到。對于8比特環(huán)境，一個包含T值的可查閱的表需要有256個記錄。</p><p>　　正如對灰度變換介紹的那樣，它顯示了圖像增強常用的三個基本類型函數(shù)：線性的(正比和反比)、對數(shù)的(對數(shù)和反對數(shù)變換)、冪次的(n次冪和n次方根變換)。正比函數(shù)式最一般的，其輸出亮度與輸入亮度可互

63、換，唯有它完全包括在圖形中。</p><p>　　3.2.1 圖像反轉(zhuǎn)</p><p>　　灰度級范圍為[0，L-1]的圖像反轉(zhuǎn)可由反轉(zhuǎn)變換獲得，表示為：</p><p><b>　　(3.2-1)</b></p><p>　　用這種方式倒轉(zhuǎn)圖像的前度產(chǎn)生圖像反轉(zhuǎn)的對等圖像。這種處理尤其適用于增強嵌入于圖像暗色區(qū)域的白色

64、或灰色細節(jié)，特別是當(dāng)黑色面積占主導(dǎo)地位時。</p><p>　　3.2.2 對數(shù)變換</p><p>　　對數(shù)變換的一般表達式為：</p><p><b>　　(3.2-2)</b></p><p>　　其中c是一個常數(shù)，并假設(shè)r≥0。此種變換使窄帶低灰度輸入圖像值映射為一寬帶輸出值。相對的是輸入灰度的高調(diào)整值?？梢岳?/p>

65、這種變換來擴展被壓縮的高值圖像中的按像素。相對的是反對數(shù)變換的調(diào)整值。</p><p>　　一般對數(shù)函數(shù)的所有曲線都能完成圖像灰度的擴散/壓縮。事實上飛，就此目的而言，下節(jié)將討論的冪次規(guī)則變換比對數(shù)變換更加靈活。不管怎樣，對數(shù)函數(shù)有它重要的特征，就是它在很大程度上壓縮了圖像像素值的動態(tài)范圍，其應(yīng)用的一個典型例子就是傅里葉頻譜，他的像素值有很大的動態(tài)范圍，這將在第四章中討論?，F(xiàn)在，我們只注意圖像頻譜的特征。頻譜值的

66、范圍從0到106過更高的情況是不常見的。當(dāng)計算機處理像這樣的無誤數(shù)字時，圖像顯示系統(tǒng)通常不能如實地再現(xiàn)如此大范圍的強度值。最后的效果是有很多的細節(jié)會在典型的傅里葉頻譜顯示時丟失。</p><p>　　3.2.3 冪次變換</p><p>　　冪次變換的基本形式為：</p><p><b>　　(3.2-3)</b></p><

67、;p>　　其中c和y為正常數(shù)。有事考慮到偏移量(即當(dāng)輸入為0是的可測量輸出)，式(3.2.3)也寫做 。不管怎樣，偏移量通常是顯示標(biāo)定的衍生，并且一般在式(3.2.3)中忽略掉。作為r的函數(shù)，s對于y的各種值繪制的曲線。如對數(shù)變換的情況一樣，冪次曲線中y的部分值把輸入窄帶暗值映射到寬帶輸出值。相反，輸入高值時也成立。然而，不想對數(shù)函數(shù)，我們注意到這里隨著y值的變換將簡單地得到一族變換曲線。如預(yù)期的一樣，我們看到其中y＞1的值和y＜

68、1的值產(chǎn)生的曲線有相反的效果。最后，我們注意式(3.2.3)當(dāng)c = y 1時，將簡化為正比變換。</p><p>　　用于圖像獲取、打印和顯示的各種裝置根據(jù)冪次規(guī)律進行響應(yīng)。習(xí)慣上，冪次等式中的指數(shù)是指伽馬值[因此在式(3.2-3)中用到這一符號]。用于修正冪次響應(yīng)現(xiàn)象的過程稱作伽馬校正。</p><p>　　如果涉及在計算機屏幕上精確顯示圖像，伽馬校正是很重要的。不恰當(dāng)?shù)膱D像修正會被漂

69、白或變得更暗。試圖精確再現(xiàn)顏色也需要伽馬校正的一些知識，這是因為改變伽馬校正值不僅可改變亮度，還可改變紅、綠、藍的比率。對于成百上千萬的網(wǎng)民(這些人的絕大多數(shù)都有不同的監(jiān)視器或監(jiān)視器設(shè)置)瀏覽的流行網(wǎng)站，為其創(chuàng)作圖像是經(jīng)常的事。有些計算機系統(tǒng)甚至配有部分伽馬校正。同時，目前的圖像標(biāo)準(zhǔn)沒有包括創(chuàng)作圖像的伽馬校正值，因此，問題更加復(fù)雜化了。由于這些限制，當(dāng)在網(wǎng)站中存儲圖像時，一個可能的方法就是用伽馬值對圖像進行預(yù)處理，此伽馬值表示了在開放的

70、市場中，在任意給定時間點，各種型號的監(jiān)視器和計算機系統(tǒng)所被期望的“平均值”。</p><p>　　3.2.4 分段線性變換函數(shù)</p><p>　　對前面三小節(jié)中所討論方法的補充是分段線性函數(shù)。其相比前面所討論函數(shù)的主要優(yōu)勢在于它的形式可任意合成。事實上，可以立刻看到，有些重要變換的實際應(yīng)用可由分段線性函數(shù)描述。分段線性函數(shù)的主要缺點是其需要更多的用戶輸入。</p><

71、p><b>　　對比拉伸</b></p><p>　　最簡單的分段線性函數(shù)之一是對比拉伸變換。低對比度圖像可由照明不足、成像傳感器動態(tài)范圍太小，甚至在圖像獲取過程中透光鏡光圈設(shè)置錯誤引起。對比拉伸的思想是提高圖像處理時灰度級的動態(tài)范圍。</p><p><b>　　灰度切割</b></p><p>　　在圖像中提高特

72、定灰度范圍的亮度通常是必要的，其應(yīng)用包括增強特征(如衛(wèi)星圖像中大量的水)和增強X射線圖中的缺陷。有許多方法可以進行灰度切割，但是，它們中的大多數(shù)是兩種基本方法的變形。其一就是在所關(guān)心的范圍內(nèi)為所有灰度指定一個較高值，而為其他灰度指定一個較低值。</p><p><b>　　位圖切割</b></p><p>　　代替提高灰度范圍的亮度，而通過對特定位提高亮度，對整幅圖像

73、質(zhì)量仍然是有貢獻的。設(shè)圖像中的每一個像素都由8比特表示，假設(shè)圖像是由8個1比特平面組成，其范圍從最低有效位的位平面0到最高有效位的位平面7。在8比特字節(jié)中，平面0包含圖像中像素的最低位，而平面7則包含最高位。</p><p><b>　　3.3直方圖處理</b></p><p>　　灰度級為[0,L-1]范圍的數(shù)字圖像的直方圖是離散函數(shù) ，這里 是第k級灰度， 使圖像

74、中灰度級為 的像素個數(shù)。經(jīng)常以圖像中像素的總數(shù)(用n表示)來除它的每一個值得到歸一化的直方圖。因此，一個歸一化的直方圖由 給出，這里 。簡單地說， 給出了灰度級為 發(fā)生的概率估計值。注意，一個歸一化的直方圖其所有部分之和應(yīng)等于1.</p><p>　　直方圖是多種空間域處理技術(shù)的基礎(chǔ)。直方圖操作能有效地用于圖像增強，如本節(jié)所示。除了提供有用的圖像統(tǒng)計資料，在以后的章節(jié)會看到直方圖固有的信息在其他圖像處理應(yīng)用中也是