Transport and Telecommunication Institute

Faculty of Computer Science and Telecommunications

COMPUTER VISION & IMAGE

PROCESSING
Study Course of Master's Programme
in Computer Science (45481)

2020 © A.Grakovski 2020

 COMPUTER VISION and IMAGE PROCESSING
About 90% of the information a person receives in the form of visual information
that must be received, perceived and analyzed. The use of equipment for
processing and analyzing video information can be both for the purpose of
converting and modifying incoming data for subsequent provision to a person
(Image Processing), and for the purpose of automatically analyzing and extracting
information, including for choosing a subsequent reaction (Computer Vision).

Fields of application: various surveillance systems, technical vision, monitoring,

video telephony, where huge amounts of video data are recorded and transmitted,
various autonomous systems (robots) that make decisions based on the analysis of
video information, television broadcasting, printing, medicine, the Internet, and
many other areas.
2
Goal of the course:
course:
To form an understanding of the tasks of computer vision and image analysis, the
most common methods for obtaining and displaying graphic images and video
information, filtering, restoration, correction and segmentation of images,
recognition and classification of objects in the image.

Learning outcome:
outcome:
Knowledge:
• current state and development trends of computer vision, basic principles and
concepts of image processing and video streams, including color models and
hardware solutions;
• main stages, approaches, algorithms for processing and converting digital
images, applications, main methods and operations of digital image processing.
Skills:
• implement in software form the methods and algorithms for digital image
processing, including preprocessing, background removal, binarization,
smoothing;
• apply in software form the methods of filtering, segmentation and recognition of
objects in digital images.
Competences:
• perform a comparative analysis and classification of methods and algorithms
for computer vision, use algorithmic tools for solving the problems of
processing and analysis of digital images.

3
 COMPUTER VISION and IMAGE PROCESSING
Main parameters of study course:
Course volume: 4 c.p. (6 ECTS)
160 academic hours (48 contact hours),
lectures – 24 hours,
laboratory and practical works – 24 hours,
independent student’s work – 112 hours.
Assessment: Laboratory works in the team – (20%),
Home work – (20%),
Final exam – (60%).
Requirements: Discipline, attendance, attentiveness,
fulfillment of tasks on time, independent creative work.

Lecturer: Alexander Grakovski

Dr.sc.ing., Professor
Electronics and Telecommunications Department
(office 606, phone +371 67109394), e-mail: [email protected]

4
 COURSE CONTENT
Part-Time,
No Themes of classroom lessons (volume, hours) Distant

Introduction. General characteristics of the current state, trends and the main
1 stages of digital image processing. 4
Methods and means of obtaining and displaying digital images. Basics of
2 digital representation of images. Discretization and quantization of images, 4
color models.
Pre-processing, background, noise, binarization. Luminance operations,
3 spatial filtering. 6
Filtering in the frequency domain. Compression, restoration and
4 reconstruction of digital images. Codecs. Watermarks. 6
Morphological image processing and cellular automata. The “deep learning”
5 approach and artificial neural networks (ANN). 4
Algorithms of segmentation. Representation and description of images.
6 Signatures of the objects. 6
Identification of objects in the image. Aerial photography and satellite
7 imagery. Fractal representation. Stereo data processing and 3D 4
reconstruction.
OCR problem. Recognition of printed and handwritten texts. Recognition of
8 car ID numbers (state plates) and bar codes. 4
9 Human biometry. Identification of the face, fingerprints, hands, iris, veins. 6
10 Recognition of moving objects in video surveillance systems. 4

Total: 48
5
 Basic literature/sources for studies:
1. Gonzalez, R.C. and Woods, R.E. Digital Image
Processing, 3rd edition, Pearson Prentice Hall, 2008,
976 p.
2. Davies, E. Machine Vision: Theory, Algorithms,
Practicalities. 3nd Edition, Elsevier, 2004, 934 p.
3. Gonzalez, R. C., Woods, R. E. and Eddins, S. L. Digital
Image Processing Using MATLAB ®, 2nd edition,
Gatesmark Publishing, 2009, 827 p.
4. Szeliski, R. Computer Vision. Algorithms and
Applications. Springer-Verlag, London, 2011, 812 p.
5. Haralick, R., Shapiro, L. Computer and Robot Vision.
Addison-Wesley Publishing Company, 1992, 673 p.
6
 Theme 1. INTRODUCTION TO IMAGE PROCESSING
An important role in the development of modern technology is played by the
methods of processing graphic information. They provide improved images
for their best perception, compression of video data for storage and
transmission through the communication channels, as well as analysis,
recognition and interpretation of visual images for decision making and
control the behavior of autonomous systems.

Within the framework of

commercial secrets, the
user frequently doesn’t
have the opportunity to
understand the actions
performed, i.e. neither to
be sure that when they
are applied to other data,
the expected result will be
obtained, nor to draw up a
correct plan of necessary
actions to achieve some
set goal.
7
 History of Digital Image Processing
1920-1929 – in newspaper business for the transmission of illustrations
over the transoceanic submarine cable between London and New York

1964 - computer processing of images of the lunar

surface transmitted by the “Ranger-7” spacecraft
(Jet Propulsion Laboratory in Pasadena)

1979 - Computed X-ray image processing

and computed tomography (Godfrey
Hounsfield and Alain Cormack)

Images in different parts of the EM spectrum

Electron UV image of the IR image Ultrasound image Radar image

microscopy (EM) material (video frame) (bone sonogram) (satellite, Tibet) 8
 Key Terms and Definitions (1)
The image is a two-dimensional function f(x, y),
where x and y are the coordinates on the plane,
and the value of f at the point (x, y) is called the
intensity or level of brightness (in gray scale).
If the values x, y and f take a finite number of
discrete values, then we are talking about a digital
image. Such points are image elements or pixels.

In contrast to the vision of people who are able to perceive electromagnetic radiation
only in the visible range, machine image processing covers almost the entire
electromagnetic spectrum from gamma radiation to radio waves. Image processing
and machine vision are related fields.

Image processing
is defined as a
discipline where
images are
present at the
input and output
of the process.

Machine vision is a computer simulation of human vision, including learning, the ability
to inference and act on the basis of observable information (imitation of human
intellectual activity).
9
 Key Terms and Definitions (2)

Image segmentation is the operation of selection on the image of areas where

objects for recognition are located and additional operations on them are
possible.

Signature of an object is the set of quantitative features that allow classification

and identification of the object. 10
 Main stages of digital image processing (1)

Image recording is the process of obtaining a digital image, it

is determined by the recording equipment introduced
distortions and shooting conditions. May contain several
stages, including sampling, information representation,
conversion and compression.

Image improvement is the process of of image manipulation, as a result of which it becomes

more suitable for a specific application than the original (the improvement methods are problem-
oriented (different for different tasks: distortion correction, filtering and noise removal, contrast,
equalizing). Image improvement is based on subjective preferences of the human perceptions.

Image recovery is also an increase in the visual quality of the image. In contrast to improvement
(criteria are subjective), image restoration is objective based on mathematical or probabilistic
models of image distortion (resampling, distortion correction, noise filtering and removal,
contrast, smoothing, sharpening).
11
 Main stages of digital image processing (2)
Image compression is the reducing the volume of memory needed to store an image, or
narrowing the channel bandwidth required for its transmission. Compression is reflected in the
names of graphic files, for example, * .jpg (Joint Photographic Experts Group - JPEG), * .gif, * .avi,
etc.
Image segmentation splits an image into its component parts or objects. Automatic segmentation
is one of the most difficult tasks of digital image processing. Overly detailed segmentation
requires a long solution path, requiring the identification of objects separately. Insufficiently
detailed or incorrect segmentation will inevitably lead to errors at the recognition stage. The more
accurate the segmentation, the greater the chances of success at recognition.

Representation and description immediately follow the segmentation stage, the output of which is
usually only data on the pixels that either form the boundary of the region (i.e., a lot of pixels are
given that separate one area of the image from another) or represent all points of the regions
themselves.
Forming of description is the selection of features (signatures), it is associated with the selection
of attributes that would express the information of interest in a quantitative form or could serve as
the basis for distinguishing classes of objects (classification) and further for identification.

Image recognition is the

process of assigning an
identifier to an object (for
example, “vehicle”) based on
a certain set of attributes
(signature).

12