POLOTSK STATE UNIVERSITY
MSc 1-40 81 01
Computer Science and Software Development
Technologies
Intelligent Image Processing
Syllabus
Dr. Rykhard Bohush
2020-2021 Academic Year
MSc 1-40 81 01
Computer Science and Software Development
Technologies
Course: Intelligent Image Processing
Faculty: ICT Amount of student effort (hours): 210
Department: Computing Class contact time (hours): 96
Systems and Networks
Level: MSc, Year 1,2 Term period(s): Sem. 2,3
Sem.2 (hours): 60
Sem.3 (hours): 36
Language of instruction: Course status: Mandatory core
English
Course leader: Entry requirements: None
Rykhard Bohush
Office: Room 158, Building “B” (Streletskiy str., 4,
Polotsk)
Learning goals and learning outcomes
Objectives:
➢ to introduce principles of image and video description for intelligent
processing in computer systems;
➢ to equip students with knowledge and skills for application two-
dimensional discrete transformations;
➢ to introduce principles and practices of methods for improving the quality
of images in spatial and frequency areas;
➢ to help students learn methods for object detection, analysis and
recognition in images and video;
ERASMUS+
➢ to equip students with knowledge and skills for application convolution
neural networks;
➢ to help students understand principles of video sequences processing
methods and algorithms for object detection and tracking;
➢ to equip students with knowledge and skills for using open computer vision
libraries;
➢ apply image processing knowledge and skills to practical matters from
different professional practices;
➢ collaborate in teams and do project-based tasks for image and video
processing;
Learning outcomes
Upon successful completion of the course, students will be able to:
➢ explain how digital images and video sequences are represented in a
computer;
➢ know how go from the mathematical description of image processing
techniques to code;
➢ apply knowledge of computing and mathematics appropriate to intelligent
image processing;
➢ analyse a problem and identifies the computing requirements appropriate
for its solution;
➢ design, implement and evaluate a computer-based system, process,
component or program to meet desired needs;
➢ apply mathematical foundations, algorithmic principles and computer
science theory to the modelling and design of computer-based systems;
➢ analyse a wide range of problems and provide solutions related to the
design of image processing systems through suitable algorithms, structures,
diagrams, and other appropriate methods;
➢ use the documentation for, and make use of OpenCV library and Deep
Learning Frameworks;
➢ write a program which implements image and video processing algorithms
for practical tasks;
➢ work as a member of a small group working on some programming tasks
for intelligent image processing;
➢ self-evaluate their projects and assess the others’ contributions and the
overall process of tasks intelligent image and video processing.
Course methodology
For successful study, the material in this course uses classical teaching
methods and didactic.
➢ Lectures
Lectures are delivered according to the published plan and on the basis of the
materials that are made available to the students.
ERASMUS+
➢ Laboratory Work
Laboratory exercises are completed according to the published plan and
based on the instructions available to students. Laboratory involves software
implementation of various methods for image processing and analysis.
➢ Consultations
Oral or On-line consultations are available on student request include Google
Classroom.
➢ Other Forms of Group and Self Study
Optionally students can work in semester on team projects. At the end of
semester, the team prepares a report about the project and presents the project to
other students and the associate professor, followed by a discussion.
Student exams are in the form of a mid-term exam and a final exam that are
graded together with other student activities. Students who do not pass through
continuing evaluation can take exam at a final examination term.
Students work collectively on case studies as a way to integrate their insights.
Students are responsible for the planning themselves and need to meet up in between
the tutorials to work on the case study.
Student presentations on selected topics. E-learning material will be available
for this course: PDF of the lectures and student presentations, and possibly audio
of the lectures. These E-lectures materials will be available on an e-Learning
platform (Google Classroom). Which will also be used for discussions with the
teacher and between the students.
Compulsory requirements:
-Oral presentation. Study one topic and present it at one of the lectures or
implement an advanced algorithm and report on it.
➢ Exam
Course Policies and Academic Honesty
Students are required to attend classes, in case classes are missed, students
should present work, which has been done during classes. Students are required to
do work at classes and online using Google Classroom and Microsoft Teams.
Submit your laboratory work covered with a sheet containing your name,
course title, number and title of the laboratory work. Any completed laboratory work
be handed in to google classroom. You must keep a duplicate copy of your work in
your PC or cloud. You should include in with your assignments: a brief report to
explain your findings; describe your algorithms and solution of tasks; a listing of
your test programs; code of your program.
ERASMUS+
Each student is expected to abide academic honesty policy. Any work
submitted by a student in this course will be the student’s own work. Students are
encouraged to study together and to discuss information and concepts covered in
lectures. However, the permissible cooperation should never involve one student
having possession of a copy of all or part of the work done by someone else.
Course Outline
Semester 2
PART 1 – IMAGE PROCESSING AND ANALYSIS
Image Representation
Reading and recording images. Discretization and quantization. Description
and presentation of binary, grayscale and colour images.
Two-dimensional discrete transforms
Fourier, Hadamard, cosine, sine, Haar
Wavelets
One-dimensional wavelet transforms. Fast wavelet transforms. Two-
dimensional wavelet transforms. Wavelet packets.
Image Enhancement Techniques
Gradational transformations. Histogram modifications. Spatial filtering.
Filtering in frequency domain.
Binary Image Morphological Processing
Binary image processing: erosion and dilatation; opening and closing.
Highlighting the boundaries. Hole filling. Isolation of connected components.
Thinning.
Gray -Scale Image Morphological Processing
Features of erosion and dilatation operations. Building transformations of
opening and closing. Basic algorithms for grayscale morphology.
Colour image processing
The basics of colour theory. Colour models. Colour conversions. Smoothing
and sharpening. Segmentation of colour images.
Segmentation of points and contour lines
Detection of isolated dots and lines. Difference models. Methods for detecting
contour differences. Linking contours and finding boundaries.
Segmentation by thresholding
ERASMUS+
Global Threshold Processing. Optimal method global threshold conversion.
Multiple Threshold Processing. Variable Threshold Processing.
Region based segmentation
Growing region. Separation and merging of region.
Segmentation by morphological watersheds. Segmentation Algorithm.
Image representation and description
Chain codes. Fourier descriptors. Moments. Signatures Border descriptors.
Area descriptors. Using the main components to describe.
PART 2 – OBJECT DETECTION AND RECOGNITION
Object Detection
Detection using keywords. Features of colour and texture similarity. Similarity
features of object shape and their location on the image.
Pattern Recognition Methods
The main tasks of pattern recognition. Recognition errors. Recognition by a set
of measurements. Structural recognition methods. Matrix of inaccuracies. Decision
tree. Bayesian decision making method. The concept of clustering methods.
Neural networks
Model of an artificial neuron. Classification of neural networks. Activation
functions. Hopfield Neural Network. Multilayer perceptron.
Deep Learning Methods for Intelligent Image Processing
The general structure of the convolutional layer. Input and output data of the
convolutional network. Error back propagation method for convolutional neural
networks. Determining the number of trained parameters. Estimating the amount of
memory needed to store a network.
CNN architectures
Features and model of the convolutional neural network SNS LeNet 5. The
architecture of AlexNet and its differences. ResNet, YOLO CNN family
Semester 3
PART 3 - VIDEO SEQUENCES PROCESSING
Motion Estimation is one of the most effective ways to utilize time redundancy
in video frames, and this part discusses various fundamental ways to do this - optical
ERASMUS+
flow, block-based methods, Bayesian methods and CNN approaches. Each of ways
are related mathematical tools which is popular and useful for students to learn for
their research.
Computer Vision and Deep Learning Libraries
Purpose and main features of OpenCV. Build and install the OpenCV library.
Using the library in Microsoft Visual Studio. Build and install the dlib library. Deep
learning image processing algorithms in dlib. TensorFlow, PyTorch, Keras.
Computer Vision Libraries for video processing
Key components: capture, play and save video. Deep Learning methods in
OpenCV and dlib
Background subtraction
Approaches using time difference methods. Building on the basis of
background subtraction methods (MOG, MOG2, ViBe). Pyramid representation of
frames. Optical flow methods.
Moving objects segmentation and classification
Segmentation of objects in video sequences. Postprocessing. Use of
classification metrics and temporal consistency. Methods based on the analysis of
movement.
Object tracking algorithms
Detection of singular points. Tracking points. Track region containing the
image. Tracking methods that consider a set of parameters. Template methods of
optical flow.
Deep learning video processing
Deep learning video processing algorithms. Using OpenCV, dlib, TensorFlow,
PyTorch, Keras for practical tasks.
Assessment Strategies:
The final grade will be based on individual grades received on assignments,
group projects and participation. Point values for assignments are based on the
level of effort and knowledge required to complete each.
The course grade is determined as following
Case studies 20 %
Active part in discussions 10 %
Laboratory works -30 %
Final exam - 40 %
ERASMUS+
Total 100%
SYLLABUS
FALL 2012
Readings, materials, and resources
• Periodicals from International Journals and Conferences
• Gonzalez, Rafael C., and Richard E. Woods. Digital image processing, 4th
Edition, 2018,
• Gary Bradski, Adrian Kaehler Learning OpenCV, 2018
• Computer Vision: Algorithms and Applications - http://szeliski.org/Book/
• OPENCV –https://opencv.org
• DLIB - http://dlib.net
• PYTORCH - https://pytorch.org/
• KERAS - https://keras.io
ERASMUS+
