Miszalok: Image Processing: Introduction

	Image Processing: Introduction Copyright © by V. Miszalok, last update: 2011-01-26
Let me know what you think	Fundamental Definitions Image Processing and Other Branches of Computer Science

Fundamental Definitions

following Rosenfeld, Kak 1982

Image	Conversion of a real world scene into a flat illumination pattern being a function f(x,y) of two variables x∈IR and y∈IR, where x is the horizontal and y the vertical coordinate in a finite, rectangular, flat region. The function values are nonnegative bound brightnesses.
Discretization	Projection of an image on an M x N lattice and sampling it at an M x N array of points.
Quantization	Mapping the continuous range of brightnesses by a finite set of integer values such as 0 ... 255.
Binarization	Mapping the continuous range of brightnesses by two integer values such as 0 and 1.
Digital Image	Finite M x N array resulting from 1. discretization and 2. quantization.
Binary Image	Finite M x N array resulting from 1. discretization and 2. binarization.
Image Processing	Manipulates and analyzes digital images by computer. Synonyms: digital imaging, picture processing
Raster Graphics	Is a set of technologies and data structures aimed to represent digital images.

Image Processing and Other Branches of Computer Science

There is an important difference between digital images and usual computer data.
Digital images usualy contain an enormous amount of redundancy and interference.
They are directed at human eyes and the human visual cortex which are specialized in consuming heavily disturbed data streams.

See: www.miszalok.de//Lectures/L01_Redundancy/Redundancy.htm

Modern Computer Science since 1980

There are four types of how to handle REDUNDANCY in computer science:

Data Processing - converting numbers to numbers
Transforming from a low-REDUNDANCY code to another low-REDUNDANCY code
Examples: book keeping, administration, statistics
Computer Graphics - converting numbers into pictures
Transforming from a low-REDUNDANCY code to high-REDUNDANCY code
Examples: Graphical user interfaces such as MacOS, Windows, Linux; CAD, games and cartoons, visualizing tools
Image Processing - changing pictures to pictures
Transforming from a high-REDUNDANCY code to another high-REDUNDANCY code
Examples: Scanners, copiers, Digital video
Computer Vision - understanding images
Transforming from a high-REDUNDANCY code to a low-REDUNDANCY code by destroying REDUNDANCY
Examples: Bar code reader, OCR, Fingerprint recognition, Chromosome- & Tumor recognition, Friend-or-Foe recognition etc.

	Caution:
1.	Do not confuse image processing IP with computer vision CV ! Both work with digital images but IP mostly generates images from images whereas CV eliminates redundancy and generates just a few measured values.
2.	Do not confuse computer vision CV with image compression IC ! Both reduce redundancy but CV doesn't preserve any pictorial information whereas IC is intended to preserve the image as good as possible.

Two Sources of Digital Images: Cameras and Graphics Cards

Nearly never digital images are composed by a human being,
because there is no economic way to select and arrange thousands or millions of pixels by hand.
(Rare exceptions: The antique, medieval and renaissance mosaics are sorts of hand-made raster graphics.)
Nowadays almost always a digital image comes out of a camera or a graphics card.

There are two main sources of digital images:
1. Scanning the real world:
digital photo, copier, TV, x-ray, CT, ultrasound, radar, weather satellite ...
2. Scanning vector graphics:
A graphics processor (= rasterizer) converts vector graphics into "Raster Graphics".

Home

Contents

Next >>

Page Top