UNESCO Module: Introduction To Computer Vision And Image Processing

Content of the courseChapter 1: Image presentation Chapter 2: Statistic operations Chapter 3: Spatial operations and transformations Chapter 4: Segmentation and edge detection Chapter 5:

Trang 1

UNESCO module:

Introduction to Computer Vision

and Image Processing

Department of Pattern Recognition and Knowledge Engineering

Institute of Information Technology

Hanoi, Vietnam Represented by LUONG CHI MAI

lcmai@ioit.ncst.ac.vn

Trang 2

Outline of the presentation

This presentation summarizes the content and organization

of lectures in module Image Processing and Computer

to Lectures

Discussion and

Conclusion

Trang 3

The course provides fundamental techniques of Image Processing and Computer Vision as well issues in practical use.

Trang 4

computers is necessary,

language will enhance the usefulness of the algorithms used in programming,

theory is helpful in mastering

transforms and compression.

Trang 5

Target audience

specialists, multimedia developers, and imaging professionals will all

appreciate Computer Vision and

Image Processing's solid introduction

Trang 6

What’s the Image Processing?

Trang 7

 Computer Vision (CV): to create a model of the

real word from images A CV system recovers

useful information about a scene from its

two-dimensional projections This recover requires the inversion of a many-to- one mapping

Vision:=Geometry+Measurement+Interpretation

What’s Computer Vision ?

Trang 8

Relationships between subjects (1)

Many fields are related to Computer Vision

Image Processing (IP): techniques usually transform images into

other images, (enhancement, correcting blurred, out-of-focus,

compression  better 2D projection image for CV).The task of information recovery is left to human user

Computer Graphics (CG): generates images from geometric

primitives such as lines, circles, and free-form surfaces CV is

the inverse problem: estimating the geometric primitives and

other features from images

CG: Synthesis of images

CV: Analysis of images

Trang 9

Relationships between subjects (2)

Pattern Recognition (PR): classifies numerical and symbolic

data Techniques: statistical and syntactical PR techniques play

an important role in CV for recognizing objects Object

recognition in CV usually requires many other techniques

Artificial Intelligence (AI): is concerned with designing systems that are intelligent and with studying computational aspects of intelligent CV is often considered as a sub-field of AI

Psyochophysics: along with cognitive science, studies human

vision for a long time Many techniques in CV are related to

what is known abut human vision

Trang 10

Content of the course

Chapter 1: Image presentation

Chapter 2: Statistic operations

Chapter 3: Spatial operations and transformations

Chapter 4: Segmentation and edge detection

Chapter 5: Morphological and other area area operations Chapter 6: Finding basic shapes

Chapter 7: Reasoning, facts and inference

Chapter 8: Pattern recognition and training

Chapter 9: Frequency domain

Chapter 10: Image compression

Trang 11

About the Chapters

Chapters

 1, 2, 3, 4, 5, 9, 10 related to Image Processing:

well known techniques to enhancement images.

 6, 7, 8 related to Computer Visions

Trang 13

Image presentation (2)

 1.2 Color representation:

Color systems: RGB, CMY/CMYK, HSI, YC b C r

Trang 14

Trang 15

- Gamma correction function

- Contrast streching End-in-search2.2 Histogram equalization

Trang 17

Trang 18

Spatial operations and

transformations (1)

Combining the techniques and operations that deal with pixels and their neighbors (spatial operations)

- Spatial filters (normally removing noise by reference to the

neighboring pixel values),

- Weighted averaging of pixel areas (convolutions),

- Comparing areas on an image with known pixel area shapes so as to find shapes in images (correlation)

- Edge detection and on detection of "interest point"

Trang 19

1 0

) ,

( ) , ( )

X I T

I(x,y) - image

T(i,j) - template of the size n x m

Trang 20

Trang 21

transformations (4)

3.4 Two dimensional geometric transformations

Frequently it is useful to zoom in on a part of an image, rotate, shift, skew or zoom out from an image

If (x’,y’) - the new coordinates and (x, y) - original coordinates

Trang 22

Trang 23

Segmentation and edge detection (1)

 Segmentation: basic requirement for the identification and

classification of objects in scene

 Techniques: splitting an image up into segments (also call regions

or areas), each holds some property distinct from their neighbor

 Approaches :

- identifying the edges (or lines) that run through an image

- identifying regions (or areas) within an image

Region operations is the dual of edge operations Ideally edge and region operations should give the same segmentation result,

however, in practice the two rarely correspond.

Trang 24

Trang 25

 4.3 First order derivative for edge detection

Hc = y_differ(x, y) = value(x, y) – value(x, y+1)

Hr = X_differ(x, y) = value(x, y) – value(x-1, y)

 4.3 Second-order edge detection

 4.4 Pyramid edge detection

 4.5 Crack edge detection

 4.6 Edge following

Trang 26

Chapter 5: Morphological and other area area operations

Chapter 6: Finding basic shapes

Trang 27

Morphological and other area

Trang 28

Morphological operations (2)

5.2 Basic morphological operations

– Binary dilation

– Binary erosion

 5.3 Opening and closing operators

Example: The use of opening: (a) An image having many connected objects, (b) Objects can be isolated by opening using the simple structuring element, (c) An image that has been subjected to noise, (d) The noisy image after opening showing that the black noise pixels have been removed.

Trang 29

Chapter 5: Morphological and other area area operations

Chapter 6: Finding basic shapes

Trang 30

Finding basic shapes (1)

 Previous chapters dealt with purely statistical and spatial

operations

 Techniques:

- looking at and processing whole images

- uses information generated by the algorithms in the previous chapter

- finding basic two-dimensional shapes or elements of shapes by

putting edges together to form lines that are likely represent real edges.

Trang 31

Finding basic shapes (2)

 6.2 Hough transforms

 6.3 Bresenham’s

algorithms

 6.4 Using interest point

 6.5 Labeling lines and

Shotest distance from origin to line defines the line in term of r and 



x y

Four cicles coincide here only

Trang 32

Trang 33

Reasoning, facts and inference (1)

- Moving from the standard IP approach to CV to make

statement about the geometry of objects and allocate labels to

them

- Enhancing by making reasoned statements, by codifying facts,

and making judgments based on past experience

- Introducing to some concepts in logical reasoning that relate

specifically to CV

- Introducing training aspects of reasoning systems The

reasoning is the highest level of CV processing

Trang 34

- Constructing a set of facts

- Constructing a rule base.

 7.2 Strategic learning

Example: A pedestal training and a pedestal description

Trang 35

7.3 Networks and spatial

– P with the visual property or

– R at this position with respect to

7.4 Rule orders

Shyni Top

Above

Table Legs

Leg

P R

R

L

L C

Trang 36

Trang 37

Pattern recognition and training (1)

 Previous chapter presented some methods used in reasoning about facts from image: edges or textures, colours or surface positions

 Some problems are better described as problems of determining a high level fact from a pattern of some kind The term "pattern" has

a wide range of meanings,

 We are particularly interested in sets of value that describe things, normally where the set of values is of a known size This is

different to looking at a scene of a flat surfaced object where we do not know how many corners there are, how many edges or how

many surfaces

Trang 38

Pattern recognition and

training (2)

 8.1 General problem

Make a series of

measurements

to give a set

of values

Determine which object this set of measurements suggests is in the image Image

x1

xn

M A X M U M

O1

On

object =

Decision function generator

Decision making process

Pattern vector Score vector(highest object score

is choosen)

Trang 39

training (3)

 8.2 Approaches to the decision making process

 8.3 Decision functions

 8.4 Determining decision functions

 8.5 Non-linear decision functions

 8.6 Using cluster means

 8.7 Supervised and unsupervised learning

- Statistical: Bayesian likelihood supervised learning

- Syntactical learning.

Trang 40

training (4)

 8.4 Determining decision function:

- Searching for islands of simplicity,

- Distance or similarity measure,

A

G ro u p

B C

Trang 41

Trang 42

The frequency domain (1)

 Most signal processing is done in a mathematical space

known as the frequency domain

 In order to represent data in the frequency domain, some

- The corners have lower frequencies Low spatial

frequencies are noted by large areas of nearly constant values

Trang 43

The frequency domain (2)

Fourier Transform of a spot: (a) original image;

(b) Fourier Transform

 9.1 The Harley transform

 9.2 The Fourier transform

Trang 44

Trang 45

 Compression of images: problem of storing them in a form that systems need to get the following benefits:

- speedily operation (both compression and unpacking),

- significant reduction in required memory, no significant loss of quality in the image,

- format of output suitable for transfer or storage

Each of this depends on the user and the application.

Image Compression (1)

Trang 46

A typical data compression system.

Image Compression (2)

Trang 47

 Run Length Encoding

Trang 48

 Focus to recovering from 2D projection to create a object model:

- Coordinate system and camera calibration

- Curve and surfaces

- Dynamic vision

 Object recognition

Conclusion

Định dạng
Số trang	48
Dung lượng	1,49 MB