數(shù)字圖像處理課件(岡薩雷斯第三版)英文翻譯課件

《數(shù)字圖像處理課件(岡薩雷斯第三版)英文翻譯課件》由會(huì)員分享，可在線閱讀，更多相關(guān)《數(shù)字圖像處理課件(岡薩雷斯第三版)英文翻譯課件（105頁珍藏版）》請(qǐng)?jiān)谘b配圖網(wǎng)上搜索。

1、 Digital Image ProcessingTianjin University of Technology and EducationSchool of Electronic Engineering2017 SynopsisThe contents of the eight weeks are as follows:l The first、 second chapter is the introduction and the basic concept of image processing.l The third, fourth chapter is space domain and

2、 frequency domain transforml The fifth, eighth chapter is image enhancement and image restoration.l The sixth chapter is image codingl The seventh, ninth chapter is image segmentation and morphologyl The tenth chapter, other The first chapter is introduction1.1 From image to image engineering Image

3、and digital image Image technology and image engineering Related disciplines and fields.1.2 Image processing and analysis Image processing and analysis system Image acquisition, display, storage, communication, processing and analysis module image processingimage analysisImage understandingAbstract

4、degree Data quantityImage foundation 1.1.1 Image and digital image What is an image? An image is a visual representation of pictures, animations, etc. The difference between graphics and computer graphics is that computer graphics are from building mathematical models to generating graphics, and ima

5、ges usually refer to graphs generated from outside. The objective world is a three-dimensional space, but the general image is two-dimensional. Two dimensional images inevitably lose part of the information in the process of reflecting the three-dimensional world. Even recorded information can be di

6、storted and even difficult to recognize objects. Therefore, it is necessary to recover and reconstruct information from images, and to analyze and extract mathematical models of images so that people can have a correct and profound understanding of what is recorded in the image. This process becomes

7、 the process of image processing.1.1 From image to image engineering Why do you need digital images? Ordinary images contain enormous amounts of information and require the use of computers to process images. Therefore, it is necessary to transform ordinary images into digital images that the comput

8、er can process. Todays digital cameras can directly convert visual images into digital images. A digital image, similar to raster graphics, consisting of finite rows and finite columns. Each basic unit is called a pixel. The voxel of a three-dimensional image is also called voxel. The two-dimensiona

9、l digital image usually is a rectangle, you can use a two-dimensional array of I (x, y) to said, where x, y coordinates in a two-dimensional space is a coordinate system, I (x, y) said in the image gray value at the point and other properties. Color can be prepared three RGB monochromatic gray value

10、. Generally speaking, these coordinates and gray values are real numbers, not only depend on the selection of coordinate system, but also depend on the measurement unit of gray value. However, a digital computer can only represent finite numbers of finite word lengths. Therefore, the gray value must

11、 be discretized. Simply put, a digital image is equivalent to a finite matrix of integer values. Digital image is the object of digital image processing and analysis. The image on the left is the image processing technique . Used to test computer algorithms A standard image of actual effects . The n

12、ame of this image is lenna . It is made up of a set of numbers. Original image The width and height are 256 pixels each .There are eight bits in pixels. It is in BMP form at About 66K bytes in size. A brief history of digital image processing The generation of digital pictures is far ahead of the co

13、mputer. The first digital image transmitted by telegraph. In six and 70s, with the development of computer hardware and the discovery of fast Fu Liye transform algorithm, it was possible to use computer to process images. Since 80s, three dimensional images have been processed. Since 90s, with the i

14、mprovement of computer performance and extensive use, image processing technology has been involved in every corner of society. The image has gradually dominated the media and produced many new industries and new opportunities. The future development of the image processing is limitless. Digital ima

15、ge processing belongs to computer science, but 90% of it relies on mathematics. From this point of view, the digital image processing technology is a very ideal direction for the students of this specialty.1.1.2 Image technology and image engineering Which belongs to image technology? Image technolo

16、gy is a general term for the technology concerned with images. It is a kind of comprehensive technical engineering. It includes image acquisition, acquisition, encoding, storage and transmission, image generation, display and output, image transformation, enhancement, restoration and reconstruction,

17、 image segmentation, target detection, expression and description, feature extraction, image classification, image recognition, modeling and matching, image and scene understanding. Narrow sense digital image processing refers to the enhancement, restoration and reconstruction of an image. The objec

18、t of operation is the pixel of an image, and the output is an image. What is image engineering? (generalized digital image processing) It is composed of three systems: image processing, image analysis and image understanding. Image processing includes image acquisition and transformation from image

19、to image, so as to improve the subjective visual effect and to do preliminary processing for image analysis and image understanding. Image analysis is to extract the data of interest from the image to describe the characteristics of the target in the image. Image understanding is based on image anal

20、ysis to study the nature and relationship of each object, in order to get the understanding of the content of the image and explain the original scene. Image processing, image analysis and image understanding are three different abstract processes from low to high. This course focuses on image proce

21、ssing and analysis systems. image processingimage understandingimage analysisAbstract degree Data quantityImageDataSymbol 1.1.3 Related disciplines and fields Image engineering is a systematic study of various image theories, techniques and applications. From the research method, and mathematics, ph

22、ysics, psychology, biology, electronics, computer science can learn from each other, from the scope of it, and pattern recognition, computer vision, computer graphics and other disciplines. 1.2 Image processing and analysis1.2.1 Image processing and analysis system The image processing and analysis

23、system includes the following modules: image acquisition module, image display module, image storage module, image communication module and image processing and analysis module.1.2.2 Modular Image acquisition module Image display module Image storage module Image communication module Image processin

24、g and analysis module 1.2.1 The data structure of an image file The basic functions of a complete image processing program are: open image files, display images, image files for the specified processing, storage of image files. Since image files are large, they usually need to be compressed before s

25、torage. So opening and storing image files involves the format of the file. Format of image file The image file refers to the file containing the image data. In addition to the image data itself, the document generally has the image description information for the image to read and display. Represen

26、ts a vector or raster form of an image.In vector form, the image is represented by a series of line segments or line segments, and the gray levels of the segments can be different. Each part of the composite body can be filled with different gray levels. A vector form file has a series of commands a

27、nd data, and the result is an image. Image data file is mainly raster form, that is, the image is a collection of image points, more suitable for complex images change. Its main drawback is the lack of links between objects and pixels, and changes in the image during the expansion of the image. For

28、example, common image file types are BMP, JPG, and so on. The image processing program must consider the format of the image file, otherwise it is not possible to open and save the image file properly.Pgm format Many universities in the United States use pgm format to avoid using compressed file for

29、mats, which is very convenient for beginners. Here is an image of this format. This is a color picture in pgm format This is the pgm format, color photos, 16 part system code. The original code is not a string of branches. Its written in the form of a branch (attention： LF= SP= newline; space; #= co

30、mment line)：0 x50 0 x35 0 x0A expressP5 (LF);0 x23 0 x20 0 x20 0 x49 0 x0A expess #(SP)(SP)I(LF) ; 0 x36 0 x34 0 x30 0 x20 0 x34 0 x38 0 x30 0 x0A express 640(SP)480(LF); 0 x32 0 x35 0 x35 0 x0A expess 255(LF)0 x27 0 x27 express 23, 23,(Pixel gray value) The decoding of this image file:P5# Imported

31、from SUN image: LEGGO_HOUSE_1.0.intensity640 4802550 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x26 0 x27 0 x27 0 x27 0 x28 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x26 0 x27 0 x26 0 x27 0 x28 0 x27 0 x

32、27 0 x26 0 x27 0 x27 0 x27 0 x28 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x28 0 x28 0 x27 0 x28 0 x29 0 x28 0 x27 0 x28 0 x28 0 x28 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x27 0 x28 0 x27 0 x28 0 x28 0 x28 0 x28 0 x28 0 x28 0 x28 0 x28 0 x28 0 x29 0 x29 0 x29 0 x28 0 x28 0 x28 0 x28. PGM format digital i

33、mage file is a common form of image processing, teaching and research in American computer science. Although the volume of files is relatively large, the pixel is directly related to the number, so it is easy to check and modify. It consists of two parts:1、 the first part is the file head, which con

34、sists of several rows: The first line illustrates the type of file, for example, P2 (for black and white images) or P5 (for color images); Then is to # at the beginning of the annotation, the comment line does not perform in the software when the image is opened, can not comment, or how to # at the

35、beginning of the comment line; The first line after the comment line specifies the size of the digital image, for example, 640480 (640 pixels wide, 480 pixels high, with a space in the middle); The next line specifies the gray scale of the image, for example, 255. There are no punctuation marks at t

36、he end of each line.2、 The 2 and second parts are bitmap arrays that are no longer segmented. For example, in the 256 gray scale, the P2 type is one pixel, one byte (8 bits), and the P5 type is one pixel, three bytes (R, G, B, 8 bits each). The second chapter is about image and vision2.1 Introductio

37、n and reviewThe foundation includes three parts: vision base, imaging base and image foundation:l Visual basis (human eyes and brightness, vision, color, vision)l Imaging Fundamentals (model, geometry, and sampling quantization)l Image basis (inter pixel relation, image operation and image coordinat

38、e transformation)Be careful： Visual: refers to the stimulation of light to the senses and the sense of the visual system.Visual perception: how to form an image of the external world through vision. 2.2 Human eye and brightness vision2.2.1 Human eye image The human eye is a complex visual organ. The

39、 front of the eye has a lens, the lens of a camera. The posterior part of the eyeball has the retina. There are two photoreceptors on the surface of the retina: cones and column cells. There are six and seven million cone cells in the eye, which are sensitive to bright light and color. The eye divid

40、es the details with the aid of cones, because each cone is connected to the nerve ending. The vision of cones is called light vision. There are 750 thousand to 1 million 500 thousand column cells in the eye. They are large, and several column cells are attached to one nerve end. Low resolution, main

41、ly providing overall visual impression. Although they are not sensitive to color, they are sensitive to weak light. The vision of cylindrical cells is called dark vision.The center of the retina is the fovea, an area of about 1.5 1.5mm2 long, the cone cell density reached 150000 /mm2, is the most se

42、nsitive eye area. Human eye structure 人 眼 成 像Density and distribution of cones and column cells 2.2.2 Brightness adaptation and differentiationSubjective brightness and subjective adaptability, because the digital image displays the image with objective brightness, the human eye gets the vision with

43、 subjective brightness, so the brightness adaptability of human eye affects the result of image processing. The visual system of the human eye can adapt to the brightness level of the light, from visible darkness to glare, the difference can reach 1010 levels. But the human eye does not see objects

44、in such a large range at the same time, and can only adapt to a small range of brightness changes at the same time (level 106). lUnder certain conditions, the current sensitivity of a visual system is called the brightness adaptation level. The sensitivity is verified by experiments. In the experime

45、nt, gradually increase the light intensity of I, change a lot for I, to achieve a number of observers can perceive, when people perceived half of the increase in the I/I Weber ratio has become a lot, as the sensitivity of the current visual system. In very strong light, it is necessary to change the

46、 intensity of the light to allow multiple observers to perceive; on the contrary, in a certain intensity of light, the slight changes in light intensity will make people perceive. The subjective brightness of the human eye is nonlinear, for example, the Mach band Optical illusions of human eyes 2.3

47、color vision According to the structure of the human eye, all colors are different combinations of three basic colors All colors can be regarded as the superposition of three basic colors, can be seen as the three color (a color removal from white) superposition Three basic features of color: gray,

48、hue and saturation, and the latter are called chromaticity. The gray scale is proportional to the reflectivity of the object; the hue is related to the main spectrum in the light; the saturation is related to the purity of the hue. The three stimuli that make up the color C are X, Y, Z, and the perc

49、entages are x, y, ZX = X/ (X+Y+Z), y = Y/ (X+Y+Z), z = 1 - X - Y. 2.5 imaging transformationLet W (X, Y, Z) be the coordinates of any point in the 3D space, Z =According to the properties of similar triangles,x/ = X/(-Z), y/ = Y/(-Z),x = X/(-Z), y = Y/(-Z)Make use of point wh homogeneous coordinates

50、 (kX, kY, kZ, K) 1 0 0 0P= 0 1 0 00 0 1 00 0 -1/ 1c h = Pwh = P(kX, kY, kZ, k)=(kX, kY, kZ, -(k/)Z+k） = (x,y,z,1)x = kX/(-k/Z+k)=X/(-Z), the same way to get y and Z x Z (X,Y,Z)0 wh = P-1ch 1 0 0 0P-1= 0 1 0 0 0 0 1 0 0 0 1/ 1wh = P-1 (x,y,0,1) = (x,y,0,1). There is no coordinate component Z, so the

51、point of 3D cannot be representedIf additional depth information Z is known, then the point of 3D can be represented,(X,Y,Z,1) = w h = P-1(x,y,z,1) = (x,y,z, z/+1). 其 中 ，X = x/(+z), Y = y/(+z), Z = z/(+z) 2.6 sampling and quantification An image needs to be discretized to become a digital image befo

52、re it can be processed by a computer. The discretization of spatial coordinates of images is called spatial sampling, and the discretization of grayscale is called grayscale quantization. Sampling is divided into uniform sampling and quantification and nonuniform sampling and quantification. Suppose

53、 the image is a rectangle. Take M long N the same as the size of the grid in the plane, and the gray level is divided into G. The closest gray number at each point in each grid is taken as the gray level of the grid. Usually, take M=2m, N=2n, and G=2k. The number of bits needed to store an image is

54、equal to b=MNk. For example, an image of 128 long 128, 64 gray level needs 220, 512 long 512, 256 image gray level to 226. The number of sampling and the selection of gray level are related to the resolution and storage ability, which need to be considered comprehensively. Such as: The effect of cha

55、nges in image spatial resolution. The effect of changes in image gray scale resolution. The effect of simultaneous changes in image space and gray scale resolution. The effect of changes in image spatial resolution The effect of changes in image gray scale resolution Effect of changes in image gray

56、scale resolution (2) Nonuniform sampling and quantizationGiven the spatial resolution, the quality of the image can be improved by adaptive sampling process according to the image characteristics. For example, areas with large variations in gray levels should be sampled more closely. Also, for examp

57、le, the frequency of all gray values can be calculated. If a range of gray values appear frequently, and in other areas, the gray values appear more sparse, then in this range, the quantization gray scale is more dense, and in other areas less dilute. 2.7 Pixel connection Pixels have four neighborho

58、od N4 and eight neighborhood N8。 Pixel connectivity consists of 4- connections, 8- connections, and hybrid connections (m- connections). Call points R and P are mixed connections, if R and P are 4- connections, or R and P are 8- connected, but not 4- connected. The hybrid connection avoids ambiguity

59、 arising from 8- connectivity. 2.8 Arithmetic and logic The four operations of the image are done one by one as image points. The addition is used to remove noise commonly used in medical imaging, subtraction, multiplication and division is often used to correct the image of the gray shadow. Commonl

60、y used logical operations are（ 1） （ AND）（ 2） （ OR）（ 3） （ NOT) The above operations can be used either for the whole image or for neighborhood operations. For example, the arithmetic mean of one point and its 8- neighborhood is used as the new value of this point. Now, the unit ALU of arithmetic and

61、logic operations is used to speed up operations. The third chapter is the spatial relation of pixels Images are made up of pixels. An image is constructed in pixels. Usually x, axial left, Y axis down. The coordinate transformation of an image is the coordinate transformation of a pixel. The so-call

62、ed spatial transformation is the transformation of pixels from one space to another. One of the applications of coordinate transformation is to correct the geometric distortion of an image. The neighborhood of a pixel refers to the set of pixels around the pixel. A pixel P is surrounded by eight pix

63、els, which together form the 8- neighborhood N8 (P) of the pixel. The 4- neighborhood N4 (P) consists of one pixel and four pixels in the upper and lower parts. A pixel consists of four diagonal pixels called diagonal neighborhood ND (P). Adjacency, connection and connectivity between pixels of the

64、same gray scale: The two pixels fall into each others 4- neighborhood, called the 4- connection The two pixels fall into each others 8- neighborhood, called the 8- connection If two pixels or 4- connections, or not 4- connections, fall into each others diagonal neighborhood ND (P), then they are cal

65、led hybrid connections (m- connections). 3.1 pixel connection Mixed connections can avoid multipath problems caused by 8- connections. The connectivity of two sets of pixels means that they have a path within the specified neighborhood 。For example, the upper left corner and the lower right corner o

66、f the figureThe connection by 4- is disconnected;The connection by m- is connected;The 8- connection is connected, but the path is not unique. The distance between two pixels P and Q, i.e., a function that satisfies the following three conditions D： D(p,q) 0， if and only if p=q,D(p,q)=0. D(p,q)= D(q,p), D(p,q) +D(q,r) D(p,r). DE(p,q) represents the Euclidean distance between p=(x_p,y_p) and q =(x_q,y_q) DE(p,q) = (x_p x_q)2 + (y_p y_q)21/2 D4(p,q) represents the distance under the 1- norm D4(p,q