1 / 78

数字图像处理（硕） 2007-2008 学年第一学期计算机学院研究生专业课 教师：房斌， 65112784 （ O ） 13883119260 （ M ）

数字图像处理（硕） 2007-2008 学年第一学期计算机学院研究生专业课 教师：房斌， 65112784 （ O ） 13883119260 （ M ） 主楼 1723 ， fb@cqu.edu.cn 周次： 10-16 教室： A 主教 117 星期一下 5-8 节 成绩分布：作业（ 2 次） 2×10% 出勤： 10% 期末考试 70%. Digital Image Processing. 参考书目录：

myrrh
Télécharger la présentation

数字图像处理（硕） 2007-2008 学年第一学期计算机学院研究生专业课 教师：房斌， 65112784 （ O ） 13883119260 （ M ）

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. 数字图像处理（硕） 2007-2008学年第一学期计算机学院研究生专业课 教师：房斌，65112784（O）13883119260（M） 主楼1723， fb@cqu.edu.cn 周次：10-16 教室： A主教117 星期一下5-8节 成绩分布：作业（2次） 2×10% 出勤： 10% 期末考试 70%

  2. Digital Image Processing 参考书目录： • Digital Image Processing，Gonzalez RC & Woods RE，Prentice Hall，2007 • Pattern Classification & Scene Analysis， Duda RO & Hart PE ，John Wiley & Sons Inc, 1973 • Digital Image Processing Using MATLAB，Gonzalez RC，Woods RE & Eddins SL，Prentice Hall ，2003 • Matlab，the MathWork Inc.

  3. Chapter 1 Introduction

  4. 本田公司最新开发的新型机器人“阿西莫” 世界第一个机器人艺人“Ever-2 Muse”

  5. What is Digital Image Processing? • Vision is the ability to perceive and recognize geometrical objects, patterns, texture, and etc, and to associate them with some functionality or meaning. • Recognizing and associating are two different processes. • The goal of Digital Image Processing is to enable the process of recognition. • The ultimate goal of DIP is to enable a computing machine to recognize at least geometrical sizes, shapes and other objects as in human vision.

  6. What is Digital Image Processing? • DIP is not an electronic equivalent to human vision in any way. • DIP is a branch of Artificial Intelligence (AI). • An attempt to emulate human vision is called weak AI. • To exactly produce a human replica electronically is called strong AI.

  7. Original Image - Lena Lena with noise Lena with noise removal by lowpass filter

  8. Original Image - Lena Binary Lena by Otsu thresholding method Contour Lena by contour following for binary Lena

  9. Original Image - Lena Enhanced Lena by Histogram Euqalization Edge map by Robert operator Edge map by Sobel operator

  10. Applications of Digital Image Processing 1. In Flexible Manufacturing Systems: • Product Inspection • Pick-and-Place • Assembly • Vehicle Guidance 2. In Biomedical Engineering: • Analyzing Chromosome • Tomography • X-ray Analysis • Mammograms • Thermograms

  11. Medical product inspection

  12. Image analysis for chromosome deficiency identification in maize

  13. Applications of Digital Image Processing 3. In Military Areas: • Bomb Disposal • Infra-red Night Vision • Radar Image Processing • Target Identification 4. In Civilian Areas: • Telecommunications • Fire fighting • Fingerprint detection (biometrics) • Intelligent Vehicle Highway System

  14. Fingerprint Detection System

  15. Applications of Digital Image Processing 5. In Commercial Areas: • Bar Code Reader • Text Reader • Script Reader • Desk Top Publishing • Multimedia 6. In Scientific Experiments: • Fingerprint Detection • Space Exploration • Nuclear Reactor • Geographic Studies • Archaeology • Physics

  16. 汉王随身抄 晨拓阅读扫描笔

  17. A Brief Historic Review

  18. DIGITAL IMAGE PROCESSING SYSTEM OVERVIEW 1.Image Capturing System • To sample an analog input and reconstruct a digital version of the image. • CCD or video camera • Frame grabber and memory • Lighting instruments • Image is normally degraded by noisy environment, poor light conditions, incorrect camera calibration and quantization error. 2. Image Enhancement System • To improve the fidelity of the image before feature extraction • Brightness, contrast, sharpness, smoothness, distortion, contortion.

  19. DIGITAL IMAGE PROCESSING SYSTEM OVERVIEW 3. Feature Extraction System • Extracts the features of the objects contained in the image • Main features point, lines, edges or regions. • Other features are light intensity, histogram, textures, color. 4. Feature Representation and Description System • Coding of object features into a certain optimal form • Some are symbolic, and some are structural. • To reduce the amount of data processed and stored.

  20. DIGITAL IMAGE PROCESSING SYSTEM OVERVIEW 5. Object Classification System • To compare the current object with some known objects in the image database. • To learn the object if no known objects are found. • Associate a known argument or functionality to the object.

  21. Chapter 2 Digital Image Acquisition • The major purpose of this chapter is to describe how sensors produce digital images from the real objects. • The 2D digital image is an array of intensity (亮度) samples reflected from or transmitted through objects: • This image is processed by a computer program. • Often, a 2D image represents a projection of a 3D scene; this is the most common representation used in pattern recognition.

  22. Human Perception and Image Capture • Crudely speaking, the human eye is a spherical camera球形照相机with a 20mm focal length lens at the outside focusing the image on the retina 视网膜which is opposite the lens and fixed on the inside of the surface of the sphere. • The iris 虹膜controls the amount of light passing through the lens by controlling the size of the pupil 瞳孔. • Each eye has one hundred million receptor cells ‑ quite a lot compared to a typical CCD array. • The retina is unevenly populated with sensor cells. • An area near the center of the retina, called the fovea 视网膜中心凹, has a very dense concentration of color receptors, called cones 圆锥. • Away from the center, the density of cones decreases while the density of black‑white receptors, the rods, increases.

  23. Visible light X-rays Infrared 红外线 Blue Green Red Ultraviolet 紫外线 800 400 Wavelength (nanometers) 1. Sensing Light • Devices can sense and produce different types of electromagnetic radiation 辐射, such as radio waves无线电波, X‑rays,microwaves 微波, etc. • The human eyes are sensitive to radiation (light) with wavelengths ranging from roughly 400 nanometers (blue) to 800 nanometers (red).

  24. A simple model of common photography: A surface element, illuminated by a single source (the sun or a flash bulb) reflects radiation toward the camera, which senses it via chemicals on film. OBJECT Point source of illumination Surface element Surface normal N CAMERA Irradiance 发光 Radiance 光辉 Z Optical axis Surface reflectance Sensor element

  25. 2. Image Devices There are many different devices that produce digital images. They differ in the phenomena sensed as well as in their electromechanical design. • (1) CCD Cameras • The charge‑coupled device (CCD) is the most flexible and common input device for machine‑vision systems. • The CCD camera is very much like a 35mm film camera commonly used for family photos. • The tiny solid state cells convert light energy into electrical charge. • The image plane acts as a digital memory that can be read row by row by a computer input process.

  26. Read row or pixel To Frame Buffer To TV 3D Scene Lens Alternative analog TV signal Image plane Pixel Figure --- A CCD (charge‑coupled device) camera imaging a cylinder; discrete cellsconvertlight energyintoelectrical charges, which are represented as small numbers when input to a computer.

  27. Pixel Pixel Digital image is in discrete form

  28. Pixel Pixel Digital image is in discrete form

  29. Gray level 196 92 Pixel Digital image

  30. Camera CCD array Machine vision algorithm Lens Graphic display A/D 3D Scene Frame buffer Display processor Figure --- An entire computer system with both camera input and graphics output. • It is a typical system for an industrial‑vision 工业task or medical‑imaging医学 task. It is also typical for multimedia 多媒体computers. • The camera provides analog image 模拟图像. • A/D模数转化 converts the analog image to digital image. • The Frame buffer is a high speed image store for digital image. • The digital image in the frame buffer is available for display and for processing by computer algorithms.

  31. Picture Functions and Digital Images We now discuss some concepts and notation important for both the theory and programming of image‑processing operations.

  32. y yi=Real number f(xi, yi)=Real number xi=Real number x (1) Types of Images Definition-1: An analog image 模拟图像is a 2D image F(x, y) which - has infiniteprecision in spatial parametersx and y, and - infiniteprecision in intensity at each spatial point (x, y).

  33. y yi=Integer f(xi, yi)=Integer xi=Integer x Definition-2:A digital image 数字图像is a 2Dimage I[r, c] represented by a discrete 2D array of intensity samples, each of which is represented using a limited precision. • It is common to record intensity as an 8‑bit (1‑byte) number which allows values of 0 to 255. • 256 different levels is usually all the precision available • - from the sensor and • - also is usually enough to satisfy the consumer.

  34. raster oriented光栅导向uses row and column coordinates starting at [0, 0] from the top left y c F[M-1,N-1] I[0,0] I[M-1,N-1] F[0,0] x • Cartesian 笛卡尔coordinate frame with [0, 0] at the lower left I[M-1,0] r Different coordinate systems used for images

  35. Cartesian coordinate frame with [0, 0] at the image center • Relationship of pixel center point [x, y] to area element sampled in array element I[i,j] y [x0+ix,y0+jy] [W/2,H/2] x F[i,j] [0,0] F[i+1,j] [x0,y0] [-W/2,-H/2]

  36. f(x,y)=0 f(x,y)=89 f(x,y)=218 Some Definitions • Definition-3: A picture functionis a mathematical representationf(x, y) of a picture as a function of two spatial variables x and y. • x and y are real values defining points of the picture. • f(x, y) is usually also a real value defining the intensity of the picture at point (x, y). Definition-4:A gray‑scale image单色灰度图像 is a monochrome digital image f(x, y) with oneintensity valueper pixel.

  37. Definition-5: A multispectral image彩色图像 is a 2D image M[x, y], which has a vector of values at each spatial point or pixel. If the image is actually a color image, then the vector has 3 elements.

  38. ()  450nm (blue) 550nm (Green) 640nm (yellow-green) Definition-5-M:Colour vision depends on three types of cone cells, each with its own relative luminance efficiency function which measures the sensitivity of the respective type of cone cells to different frequencies of light. From these curves, we can see that the human vision system is biased towards the colour green (note the significant overlap between S1 and S2).

  39. ()  450nm (blue) 550nm (Green) 640nm (yellow-green) To mimic the way we see colour, all colour image formats are based on three components, one for each of the three primary colours: Red, Green and Blue (RGB) or some linear combinations of them. There are 100 million rod cells on a human retina (roughly equivalent to 10,000 dpi) versus 6.5 million cone cells视锥细胞.

  40. y f(x,y)=1 f(x,y)=0 x Definition-6: A binary image 二值图像 is a digital image with all pixel values 0 or 1. A B 图 像

  41. Original image Labeled image Boundaries of the extracted face region • Definition-7:A labeled image分类图像 is a digital image L[r, c] whose pixel values are symbols. • The symbol value of a pixel denotes the outcome of some decision made for that pixel.

  42. Labeled image Original image (tiger)

  43. Image Plane Size of scene element pixel pixel 3D Scene Lens (2) Image Spatial Measurementand Quantization • Definition-8:The nominal resolution 标称分辨度of a CCD sensor is the size of the scene element that images to a single pixel on the image plane. • Each pixel of a digital image represents a sample of some elemental region of the real image.

  44. Image Plane pixel 3D Scene Lens • If the pixelis projectedfrom the image planeback out to the source material in the scene, then the size of that scene element is the nominal resolution 标称分辨度of the sensor.

  45. 1010 inch2 500500 • For example, if a 10 inch square sheet of paper is imaged to form a 500500 digital image, then the nominal resolution of the sensor is 0.02 inches (10/500 = 0.02).

  46. Definition-9:The term resolutionrefers to the precision of the sensor in making measurements, but is formally defined in different ways. • If defined in real world terms, it may just be the nominal resolution, as in “the resolution of this scanner is one meter on the ground” • Or it may be in the number of line pairs per millimeter that can be resolved or distinguished in the sensed image. • Definition-10: The field of view 视野of a sensor (FOV) is the size of the scene that it can sense, for example 10 inches by 10 inches.

  47. (a) Digital image with 127 rows of 176 columns; (b) (6388) created by averaging each 22 neighborhood of (a) and replicating the average to produce a 22 average block; (c) (3144) created in same manner from (b);and (d) (1522) created in same manner from (c). Effective nominal resolutions are (127  176), (63  88), (31  44), and (15  22) respectively.

  48. Image Quantization 图像量化 • A quantizer is an Analog-to-Digital device which converts a continuousinput signal u to one of a set of discrete levels called reconstruction levelsrk. • Suppose the u lies in the range: umin uumax • and we wish to quantize u into Llevels. Then we defineL+1transition levels tk:t0 = umin<t1<……<tL-1<tL = umax • The quantization step involves mapping u to its quantized value, u*, using the rule: • Define {tk, k = 0,…,L} as a set of increasing transition or decision levels with t0 and tL as the minimum and maximum values, respectively, of u.

More Related