Courses Currently Offered in English

Course Version

Course Code:EADIR
Version Number:1
Course Name:Adaptive Image Recognition
Credit Units:4
Cost Units:30
Examination type (E-with exam):B
Grading threshold:5
Initiation semester:02Z
Person responsible:prof. dr hab. Władysław Skarbek
Description:ICT programme taught in English

Hours per week

Class typeHours

Class types: W - lecture, C - tutorial, L - laboratory, P - project



Prerequitise types: W - required, Z - recommended

Similar Courses

Course CodeNameDiscount CUDiscount ECTS
AIMAnaliza i indeksowanie multimediów22

Last Course Instances

Semester codeInstance codeLecturerInstituteMax. number of students
18LAprof. dr hab. Władysław SkarbekRE40

Thematic Classification

Class CodeClass name (in Polish)
ANGLAll Courses in English (A)


Summary (in Polish)

Wykład obejmuje teorię i zastosowania adaptacyjnych technik przetwarzania obrazów, analizy obrazów i rozpoznawania wzorców do detekcji, klasyfikacji i indeksowania obiektów w obrazie. Prezentowane są algorytmy ekstrakcji cech wykorzystujące informacje o kolorze, teksturze, ruchu i kształcie obiektów. Dyskutowane są podstawowe techniki analizy dyskryminacyjnej i techniki lokalizacji obiektów. Projekty dotyczą generycznych aplikacji, takich jak: rozpoznawanie twarzy, detekcja w obrazie tablicy rejestracyjnej samochodu, rozpoznawanie cyfr, segmentacja słów oraz wyszukiwanie obrazów na podstawie deskryptorów koloru określonych w standardzie MPEG-7.

Lecture contents
  1. Introduction (2h): block diagram for generic AIR system, modules: localization/segmentation, extraction of features, core recognition task (classification/verification/search); adaptability concept of recognition system;

  2. Generic applications - requirements (2h): for face recognition, number plate recogniser, handwritten word reader, and image search by colour temperature

  3. Visual features based on colour (2h): colour space, colour quantisation, dominant colours, colour distribution, global spatial distribution, local spatial distribution

  4. Visual features based on texture (4h): energy in channels of Fourier transform of Radon transform, regularity, dominant directions, dominant scales, edge histogram

  5. Visual features based on shape (2h): region shape by ART - Angular Radial Transform, contour shape by CSS - Curvature Scale Space

  6. Statistical features (4h): representatives of scalar quantisation (SQ) and of vector quantisation (VQ), principal components coefficients (PCA)

  7. Discrimination by proximity/similarity (4h): nearest neighbours (k-NN), subspace method, linear discrimination analysis (LDA and DLDA), fractal operator, invariant reference points (IRP) methodology

  8. Discrimination by probability (4h): Bayes classifiers, Markov nets

  9. Discrimination by membership value (2h): fuzzy classifiers, rough set classifiers

  10. Localization (4h): by directional histograms, by image filtering, by Hough transform (HT)

Project contents
  1. Face recognition by principal component analysis
  2. Number plate localization
  3. Digit recogniser by principal component analysis
  4. Handwritten word segmentation
  5. Image search client server application using selected colour descriptors

Literature (mandatory)

  1. N. Sebe, M.S. Lew: Robust Computer Vision, Kluwer, 2003
  2. B.D. Ripley: Pattern Recognition and Neural Networks, Cambrige University Press, 1996

Literature (optional)

  1. Z. Liu, J. Cai, R. Buse: Handwriting recognition, Springer, 2003
  2. S. Raudys: Statistical and Neural Classifiers, Springer, 2001

Adaptive image recognition course includes topics on theory and applications of adaptive image processing, image analysis, and pattern recognition methods for image object detection, classification, and indexing. There are presented feature extraction algorithms based on colour, texture, motion, and shape information. Several fundamental techniques for pattern discrimination and localisation are discussed. Projects refer to generic applications: face recognition, number plate detection, digit recognition, word segmentation, and image search by its MPEG-7 colour descriptors.