Advanced techniques of image processing
FEKT-NPZOAcad. year: 2019/2020
The course is focused on the theory and practice of advanced image processing techniques. Video sequence processing is included. The main areas of interest are camera model, its calibration, DFT image filtering, convolution, object recognition, biometric feature recognition (skin, face, papilar lines), epipolar geometry, stereo pair analysis, correspondence problem, spatial information extraction, dynamic programming, homography, optical flow, motion in scene tracking. The emphasis is on practical aspects of image processing, i.e. algorithmizing and implementing generally formulated tasks.
Learning outcomes of the course unit
On completion of the course, students will be able to:
- describe the basic terms of digital image processing,
- apply the principles of development of image processing algorithms using the OpenCV libraries,
- describe the techniques of intensity transformations,
- define the geometrical transformations of an image,
- use homogenous coordinates,
- interpret the Discrete Fourier Transform of an image and image filtering in the frequency domain,
- describe the terms convolution, filtering and edge detection in the spatial domain,
- describe the basics of image segmentation,
- interpret the mathematical model of the camera,
- describe panoramic image stitching,
- demonstrate the properties of epipolar geometry,
- describe the principles of biometric image identification,
- describe the principles of depth information extraction,
- describe the principle, problems and methods of optical flow estimation.
Students who will sign up for the course should be able to:
describe the basic steps of signal digitalization,
interpret the basic principle of one-dimensional function derivative,
master the basic operators and functions of the C++ language,
describe the basic operations of matrix calculus,
describe basic principles of Euclidean geometry,
practice calculations with complex numbers,
discuss the basic terms from the domain of probability and statistics.
Generally, knowledge on the level of completed bachelor studies is expected.
Recommended optional programme components
Recommended or required reading
PARKER, J. R.: Algorithms for Image Processing and Computer Vision, Wiley; Bk&CD-Rom edition 1996, ISBN: 471140562. (EN)
GONZALEZ, R. C., WOODS, R. E.: Digital Image Processing, Prentice Hall 2007, ISBN: 978-0131687288 (EN)
HLAVAC V., SONKA M., BOYLE R.: Image Processing, Analysis, and Machine Vision, Cengage Learning, 2007, ISBN 978-0495082521 (CS)
PRATA, S.: C++ Primer Plus, Addison-Wesley Professional, 2011. (CS)
FISHER, R. B.: CVonline: The Evolving, Distributed, Non-Proprietary, On-Line Compendium of Computer Vision, http://homepages.inf.ed.ac.uk/rbf/CVonline/. (EN)
Planned learning activities and teaching methods
Techning methods include lectures and practical laboratories. Course is taking advantage of e-learning (Moodle) system.
Assesment methods and criteria linked to learning outcomes
- up to 60 points for the written exam
- up to 40 points for laboratory tests
The written exam is taken at the end of semester. It contains simple tasks in the form of concrete questions as to mathematical constructions, constraints of algorithms, basic equations and principles of methods, often in the form of a fill-in quiz.
Two tests are arranged in the laboratory exercises (in the 6th and 12th weeks of the semester) each for max. 20 points. The tests contain tasks corresponding to the content of realised laboratory exercises. There is no lab-test-based qualification limit for the written exam.
Language of instruction
1. Basic terms of digital image processing, algorithmisation, OpenCV library using C++ API.
2. Intensity transformations, histogram equalization.
3. Geometric image transformations, homogenous coordinates, interpolation techniques.
4. Discreet Fourier Transform, image filtering in the frequency domain.
5. Image processing in the spatial domain, convolution, filtering, edge detection.
6. Basics of image segmentation, automatic threshold estimation, Canny edge detector, Viola-Jones object detector.
7. Pinhole camera model, parallel projection, perspective projection, translation, rotation, calibration.
8. Homography, panorama image stitching, projective transformation, cylindrical, spherical panorama, basics of automatic correspondence matching.
9. Epipolar geometry, image stereo pair, rectification.
10.Biometric image identification, PCA, face recognition.
11.Spatial information extraction, correspondence problem, disparity space, dynamic programming.
12.Optical flow, scene motion estimation, object tracking, methods by Horn-Schunck, Lucas-Kanade, Farnebäck.
The aim of the course is to make students familiar with a high-quality applicative knowledge of advanced techniques of static and dynamic image processing 2D and 3D and to enable them to master both the theory and the practise of advanced techniques of image processing.
Specification of controlled education, way of implementation and compensation for absences
Laboratory lessons are obligatory. Justified absence can be made up after prior arrangement with the instructor.