Image
Processing
COMP4421
Dr. Hao CHEN, Dept. of CSE, HKUST
Our Support Team
Instructor
Dr. Hao CHEN, Assistant Professor
RM 3517, Dept. of CSE, HKUST
Email: jhc@ust.hk
Tel: +852‐2358 8346
Teaching Assistants
Mr. Yi LIN, ylindq@connect.ust.hk
Mr. Haoxuan CHE, hche@connect.ust.hk
Mr. Zhefan RAO, zraoac@connect.ust.hk
2
Teaching Schedule
• Lecture:
• Tues and Thurs 9:00‐10:20am, Room 2504 (via lift 25‐26)
• Tutorial:
• Tues 3:00‐3:50pm, Online zoom link will be provided in announcement by TA.
• Class Dates:
• Sep 1, 2022 – Nov 30, 2022
• Lecture and Tutorial notes will be available online. Part of slides credit: Dr.
Xiaojuan Qi (HKU), Dr. Ioannis Gkioulekas (CMU), etc.
• Office Hours: by appointment.
• No lab sessions.
3
Prerequisites
Basic Math
• Calculus, linear algebra, geometry,
• Probability, statistics
• Analog/digital signal processing, etc.
Programming Skills
• C++/Matlab/Python.
4
Syllabus (tentative) Lecture Topic
Introduction
Image Fundamentals
Image Transformations
1‐12
Mode: In‐person. Image Filtering
Image Restoration
We will cover some advanced topics
Image Compression
(e.g., deep learning for medical
image analysis) if time permits. 13 Mid‐term
Morphological Image Processing
Image Segmentation
Image Registration
14‐25
Feature Extraction
Image Recognition and Deep Learning
Advanced Topics 5
Books
Reference
• Digital Image Processing, by Gonzalez and Woods, 4th Ed., Pearson,
2018. Online available: https://library.ust.hk/
• Digital Image Processing using MATLAB, by Gonzalez and Woods, Prentice
Hall, 2004.
• The Image Processing Handbook, by John C. Russ.
• Digital Image Processing, by Kenneth R. Castleman, Prentice Hall, 1996.
• Two‐dimensional Signal and Image Processing, by Jae S. Lim, Prentice
Hall, 1990.
• Computer Vision: A Modern Approach by Forsyth and Ponce, Prentice
Hall, 2003.
• First Principles of Computer Vision. Columbia University. Lectures online
available.
6
Image Processing Tools
Matlab
• We use Matlab for tutorials
https://www.mathworks.com/
• Workstations in ITSC Computer Barns
http://www.ust.hk/itsc/computerbarn/
• MATLAB software and related toolboxes are
available in the computer barns, e.g., image
processing toolbox.
7
Grading Scheme
Exam 70% (Midterm 30%, Final 40%)
Assignments 30% (3 Assignments, 10% each)
Total 100%
• Details will be released before exam.
• Assignments must be submitted by 11:59pm of the due day.
• Late assignments will incur a 20% penalty per day.
• More information about submission procedures will be given.
8
Learning Outcomes
• Understand image fundamentals. • Segment image components from
an image.
• Identify basic image enhancement
techniques. • Register images with similarity
metrics and transformations.
• Image restoration in the presence
of noise and distortion. • Represent an image using different
feature descriptors.
• Compress an image with lossless
or lossy compression methods. • Apply image processing
techniques to typical applications.
• Perform morphological image
processing.
9
Two Words: Image Processing
Image
• Fundamentals
• Image formation based on individual elements (pixels)
• Sampling and Quantization
10
Two Words: Image Processing
Processing
• Manipulation of the image data
• Geometric transformations
• Enhancement/Restoration
• Image Segmentation
• Feature extraction
• Object recognition
11
Today’s Agenda
This lecture will cover
• What is a Digital Image?
• Types of Digital Images
• Digital Image Processing
• Topics Covered in this Course
12
Digital Images
(a) Natural image (b) Satellite image (c) MRI image (d) Microscopy image
(e) Cardiac CT (f) Fetus Ultrasound (g) Infrared image (h) Portrait image
One picture is worth more than ten thousand words! 13
How do Computers See an Image?
14
How do Computers See an Image?
600
400
15
Digital Image
Digital Image
An image can be defined as a two‐dimensional function, 𝑓 𝑥, 𝑦 , where 𝑥 and 𝑦 are spatial (plane)
coordinates, 𝑓 is the intensity or gray level of the image at that point.
Pixel
Digital image is composed of a finite number of elements, called image elements or pixels.
16
Digital Image
Digital Image
An image can be defined as a two‐dimensional function, 𝑓 𝑥, 𝑦 , where 𝑥 and 𝑦 are spatial (plane)
coordinates, 𝑓 is the intensity or gray level of the image at that point.
Pixel
Digital image is composed of a finite number of elements, called image elements or pixels.
Digital image processing
Digital image processing refers to processing digital images by means of a digital computer.
Gaussian Filter
Original Result 17
Today’s Agenda
This lecture will cover
• What is a Digital Image?
• Types of Digital Images
• Digital Image Processing
• Topics Covered in this Course
18
Electromagnetic (EM) Spectrum
The most familiar digital Images are based on radiation from the EM, especially
images in the X‐ray and visual bands of the spectrum.
19
https://www.britannica.com/science/microwave‐radiation
Examples: Gamma Ray
Radio Microwave Infrared Visible Ultraviolet X‐Ray Gamma Ray
20
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Examples: X‐Ray
Radio Microwave Infrared Visible Ultraviolet X‐Ray Gamma Ray
21
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Examples: Ultraviolet
Radio Microwave Infrared Visible Ultraviolet X‐Ray Gamma Ray
Image credit: Galex, NASA/JPL‐
Cygnus Loop Caltech
Fluorescence microscopy of normal (left) and ‘smut’ corns (right)
22
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Examples: Visible
Radio Microwave Infrared Visible Ultraviolet X‐Ray Gamma Ray
Remote Sensing
Finger print
Car Plate
23
Digital Image Processing, by Gonzalez and Woods, 4th Ed., Pearson, 2018. Smart Phone
Examples: Infrared
Radio Microwave Infrared Visible Ultraviolet X‐Ray Gamma Ray
Detect temperature and potential condition Land surface temperature from remote Satellite Image of Hurricane Katrina
failures; Work in low‐light environment sensing (thermal satellite imaging)
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
https://www.phase1vision.com/blog/understanding‐the‐benefits‐of‐infrared‐imaging‐cameras 24
https://chrieke.medium.com/thermal‐imaging‐with‐satellites‐34f381856dd1
Examples: Microwave
Radio Microwave Infrared Visible Ultraviolet X‐Ray Gamma Ray
Explore the inaccessible
regions of the Earth’s surface
Radar image of Southeast Tibet 25
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Examples: Radio
Radio Microwave Infrared Visible Ultraviolet X‐Ray Gamma Ray
Magnetic resonance imaging (MRI) of knee (left) and spine (right)
26
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Other Imaging Modalities
Scanning electron microscopy Fractal images generated by
of tungsten filament following computer
thermal failure
Ultrasound images of fetus (top) and thyroid (bottom)
27
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Today’s Agenda
This lecture will cover
• What is a Digital Image?
• Types of Digital Images
• Digital Image Processing
• Topics Covered in this Course
28
General Image Processing System
29
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Digital Image Processing
Low‐level Image acquisition and formation
Image enhancement and restoration
Image colour
Image processing
Image compression
Mid‐level Image segmentation
Object recognition Image analysis:
computer vision and
High‐level Scene understanding pattern recognition
30
Applications
Image/Video Super Resolution: (Note: there is an animation in this slide)
31
Xin Tao, et al. Detail‐revealing Deep Video Super‐resolution, ICCV 2017.
Applications
Low‐light Image Enhancement:
Images captured at a low‐light environment Enhanced output images
32
Chen Chen, et al. Learning to see in the dark, CVPR 2018.
Applications
Image Demoireing: (Note: there is an animation in this slide)
33
Applications
Portrait mode in mobile phone:
Slides from https://courses.cs.washington.edu/courses/cse576/20sp/calendar/ 34
https://www.deccanchronicle.com/technology/mobiles‐and‐tabs/160318/google‐pixel‐2‐portrait‐mode‐manufacturers‐deeplabv3‐software‐android.html
Applications
Image Restoration:
35
Xintao Wang, et al. GFP‐GAN: Towards Real‐World Blind Face Restoration with Generative Facial Prior, CVPR 2021.
Applications
Image Restoration:
36
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Applications
Scientific Discovery:
Image black holes
How scientists captured the first image of a black
Slides from https://courses.cs.washington.edu/courses/cse576/20sp/calendar/
hole, 2019 37
Applications
Image‐to‐Image Translation:
38
Yunjey Choi, et al. StarGAN: Unified Generative Adversarial Networks for Multi‐Domain Image‐to‐Image Translation. CVPR 2018.
Applications
Machine Vision: Image Classification
39
Slides from https://courses.cs.washington.edu/courses/cse576/20sp/calendar/
Applications
Machine Vision: License Plate Recognition
Optical
character
classification
40
credit to Dr. George Bebis
Applications
Machine Vision: Robotics, Autonomous driving and Drones
Semantic
understanding
41
Applications
Machine Vision: Object Detection
https://medium.com/analytics‐vidhya/yolov3‐object‐detection‐in‐tensorflow‐2‐x‐8a1a104c46a8 . 42
Joseph Redmon, Santosh Divvala, Ross Girshick, Ali Farhadi, “You Only Lookl Once" IEEE International Conference on Computer Vision and pattern recognition (CVPR) , 2015.
Applications
Medical Diagnosis:
Skin Cancer Detection Breast Cancer Detection
The heart and mediastinum are
normal. The lungs are clear. There is
mild blunting of the right costophrenic.
There is no infiltrate, mass or
pneumothorax. The right internal
jugular catheter has been removed.
Lungs are clear. No pleural
effusions or pneumothoraxes.
Heart and mediastinum of
normal size and contour..
Liu, Fenglin, Shen Ge, and Xian Wu. "Competence‐based multimodal curriculum learning for medical report generation." ACL, 2021 43
Applications
Style Transfer:
https://www.tensorflow.org/tutorials/generative/style_transfer 44
Applications
Image Retrieval:
45
Other Applications…
• Consumer electronics: digital cameras, scanner, . . .
• Medical and biological uses: microscopy, radiology, surgery, . . .
• Industrial uses: inspection, metrology, defect detection, . . .
• “Smart” world: intelligent transportation system, smart home, surveillance, . . .
• Scientific uses: hyperspectral imaging, remote sensing, astronomy, . . .
Consumer electronics Medical imaging Industrial imaging Autonomous driving
46
Today’s Agenda
This lecture will cover
• What is a Digital Image?
• Types of Digital Images
• Digital Image Processing
• Topics Covered in this Course
47
Digital Image Processing
Low‐level Image acquisition and formation
Image enhancement and restoration
Image colour
Image processing
Image compression
Mid‐level Image segmentation
Object recognition Image analysis:
computer vision and
High‐level Scene understanding pattern recognition
48
Topic 1: Image Acquisition
Human visual system and image acquisition: how digital images are generated
and represented; image sampling and quantization.
Human Visual System Digital Imaging System
49
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Topic 2: Intensity Transformations
Intensity transform is to change pixel intensities based on a transformation
function, such as image negative, log transformations, gamma
transformations, piecewise linear transformation function, and histogram
processing (e.g., histogram equalization).
Gamma transformation Image enhancement
50
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Topic 3: Spatial Filtering
Spatial filtering modifies an image by replacing the value of each pixel by a
function of the pixel and its neighbours, e.g., smoothing spatial filters, box
filter, gaussian filter, and sharpening filters, etc.
Image denoise
Image sharpening
Image filtering
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
51
Topic 4: Filtering in the Frequency Domain
Frequency domain filtering: frequency domain analysis and how to realize
filters in the frequency domain with applications in image smoothing and
sharpening.
52
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Topic 5: Image Restoration
Image restoration is to improve the quality of an image based on
mathematical degradation models, e.g., noise and blur model.
53
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Topic 6: Image Compression
Image compression is to reduce the storage required to save an image or
the bandwidth to transmit it.
54
Digital Image Processing
Low‐level Image acquisition and formation
Image enhancement and restoration
Image colour
Image processing
Image compression
Mid‐level Image segmentation
Object recognition Image analysis:
computer vision and
High‐level Scene understanding pattern recognition
55
Topic 7: Morphological Processing
Morphological processing deals with tools for extracting image components
that are useful in the representation of shape, such as boundary extraction,
hole filing, and skeleton analysis.
56
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Topic 8: Image Segmentation
Image segmentation is to partition an image into its constituent parts or
objects.
57
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Topic 9: Image Registration
Image registration is the process of transforming different sets of data into
one coordinate system.
58
http://reuter.mit.edu/research/neuroimg/registration/
Topic 10: Feature Extraction
Feature extraction includes feature detection and feature description.
Features are used to represent images.
Histogram
𝑥
𝑥
Feature .
𝒙 .
extraction
.
𝑥
Feature extraction “Bag‐of‐words” description of the images
59
Topic 11: Image Recognition and Deep Learning
Image recognition is to understand the semantic meaning of the image.
Deep convolutional neural network for image recognition: building blocks,
forward, and backward propagation.
60
Digital Image Processing, by Gonzalez and Woods, Pearson, 2018.
Topic 12: Advanced Topics
Biomedical Image Analysis Deep Learning for Semantic Segmentation
61
https://medium.com/analytics‐vidhya/yolo‐object‐detection‐343a430f3b48
Where does Image Processing fit in?
Computer
Vision
Machine Image Computer
Learning Graphics
Processing
Artificial
Intelligence
62
References
• Digital Image Processing, by Gonzalez and Woods, 4th Ed., Pearson, 2018.
• First Principles of Computer Vision. Columbia University.
63