1. Digital Image Processing Lecture 1: Introduction
Prof. Charlene Tsai

2. Why digital image processing?
Image is better than any other information form for human being to perceive.
Humans are primarily visual creatures – above 90% of the information about the world (a picture is better than a thousand words)
However, vision is not intuitive for machines
projection of 3D world to 2D images => loss of information
interpretation of dynamic scenes, such as a moving camera and moving objects

3. What is digital image processing?
Image understanding, image analysis, and computer vision aim to imitate the process of human vision electronically
Image acquisition
Preprocessing
Segmentation
Representation and description
Recognition and interpretation

4. General procedures
Goal: to obtain similar effect provided by biological systems
Two-level approaches
Low level image processing. Very little knowledge about the content or semantics of images
High level image understanding. Imitating human cognition and ability to infer information contained in the image.

5. Low level image processing
Very little knowledge about the content of the images.
Data are the original images, represented as matrices of intensity values, i.e. sampling of a continuous field using a discrete grid.
Focus of this course.

6. Low level image processing
3x3 neighborhood
Origin (Ox,Oy)
Pixel Value
Pixel Region
Spacing (Sy)
Spacing (Sx)

7. Low level image processing
Image compression
Noise reduction
Edge extraction
Contrast enhancement
Segmentation
Thresholding
Morphology
Image restoration

8. Low level image processing
Image compression
Noise reduction
Edge extraction
Contrast enhancement
Segmentation
Thresholding
Morphology
Image restoration

9. Low level image processing
Image compression
Noise reduction
Edge extraction
Contrast enhancement
Segmentation
Thresholding
Morphology
Image restoration

10. Low level image processing
Image compression
Noise reduction
Edge extraction
Contrast enhancement
Segmentation
Thresholding
Morphology
Image restoration

11. Low level image processing
Image compression
Noise reduction
Edge extraction
Contrast enhancement
Segmentation
Thresholding
Morphology
Image restoration

12. Low level image processing
Image compression
Noise reduction
Edge extraction
Contrast enhancement
Segmentation
Thresholding
Morphology
Image restoration

13. Low level image processing
Image compression
Noise reduction
Edge extraction
Contrast enhancement
Segmentation
Thresholding
Morphology
Image restoration
Dilation
Erosion

14. Low level image processing
Image compression
Noise reduction
Edge extraction
Contrast enhancement
Segmentation
Thresholding
Morphology
Image restoration

15. High level image understanding
To imitate human cognition according to the information contained in the image.
Data represent knowledge about the image content, and are often in symbolic form.
Data representation is specific to the high-level goal.

16. High level image understanding
What are the high-level components?
What tasks can be achieved?
Landmarks
(bifurcation/crossover)
Traces
(vessel centerlines)

17. Applications Medicine Defense Meteorology Environmental science
Manufacture
Surveillance
Crime investigation

18. Applications: Medicine
Arrows indicating metastatic lesions. The arrows to the right in the PET image indicate spleen lesions not visualized by CT. CT provides anatomic imaging while PET is for molecular imaging (to reflect metabolic activity). Exact localization of metabolic abnormalities was only possible by image fusion.
CT and MRI are important for surgical purpose, but PET has higher detection rate for malignant tumors. CT
(computed Tomography)
PET
(Positron Emission Tomography
PET/CT

19. Applications: Meteorology
Satellite images for weather observation and prediction.

20. Applications: Environmental Science
Also satellite images. Image taken from space for purposes of monitoring environment conditions on the planet. Multi-spectral images. For assessment of factors harmful to the environment, such as pollution.

21. Applications: Manufacture
Automated visual inspection of manufactured goods. Inspection for missing parts and detection of anomalies.

22. Application: Surveillance
Car Tracking Project from CMU: Tracking cars in the surrounding road scene and then generating a "bird's eye view" of the road.
The demo is car tracking. It assumes that some process has already detected the cars. It then follows them. The demo you have is hand initialized- we tell it where to start. Other people are working on this initialization. This is why we don't track all the cars - just the ones we started the algorithm on.
Courtesy of Simon Baker:

23. Applications: Crime Investigation
Original image is a fingerprint on a knife. The enhancement of the pattern aid in the automated search of a database for potential matches.
Fingerprint enhancement

24. What are the difficulties?
Poor understanding of the human vision system
Do you see a young or an old lady?

25. What are the difficulties?
Human vision system tends to group related regions together, not odd mixture of the two alternatives.
Attending to different regions or contours initiate a change of perception
This illustrates once more that vision is an active process that attempts to make sense of incoming information.

26. What are the difficulties?
The interpretation is based heavily on prior knowledge.

27. Just some fun visual perception games
Can you count the dots?

28. More … Do you see squares?
More at

29. Example: Detection of ozone layer hole
The normal value is about 300 Dobspn units (multiple of molecules per cubic centimeter). Depletion is observed. The ozon hole is observed to open each year between sep and Nov. The hole varies in seriousness and duration.
Over the Antarctic, normal value around 300 DU

30. Class Format – Efficiency of Learning
What we read 10%
What we hear 20%
What we see 30%
What we hear + see 50%
What we say ourselves 70%
What we do ourselves 90%

31. Class Format – Efficiency of Learning
This leads to in-class discussion and quizzes.
50-minute lecture
Remaining for group discussion & in-class quiz

32. Course requirements In-class quizzes 10% 4 Homework assignments 25%
Final project 25%
Midterm exam 20%
Final exam 20%
Peer learning is encouraged
BUT, NO PLAGIARISM!!!
(20% deduction if caught)

33. Textbooks Problems in picking a good textbook: Prescribed:
Hard to find a textbook of the right level --- too easy or too hard.
Hard to find a textbook of the right price --- good books tend to be too expensive
Prescribed:
Rafael C. Gonzalez, Richard E. Woods: Digital Image Processing. Prentice Hall; 2nd edition, 2002
Other references (used in 2005):
Alasdair McAndrew: Introduction to Digital Image Processing with Matlab, 2004.

34. Programming Tools
Matlab with Image Processing Toolbox for homework exercises
MATLAB Tutorial:
MATLAB documentation:
User-contributed MATLAB IP functions:

35. More on Matlab University of Colorado Matlab Tutorials:
A decent collection of Matlab tutorials, including one focusing on image processing

36. Term project Group project of 2~3 people
I decide the format of the term project
You decide your own topic that interests you
So, starting thinking about it!!!
You may implement your project with any programming language of your preference.

37. In-class quiz Goal: to enhance learning Open-book/open-notes format
Group effort of 2~3 people to encourage discussion and peer learning

38. Looking ahead: lecture2
Image types
File format
Matlab programming.