.

Introduction

Stefano Ferrari

Università degli Studi di Milano

[email protected]

Methods for Image Processing

academic year 2022–2023

Image processing

Computer science concerns the

I representation,
I processing and
I transmission
of information.

Image processing considers in particular the information

contained in (digital) images.

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 1

.

What is an image?
Some definitions:
I a percept that arises from the eyes; an image in the visual
system [TheFreeDictionary];
I an artifact that has a similar appearance to some
subject—usually a physical object or a person. [Wikipedia];
I a reproduction or imitation of the form of a person or thing
[Merriam-Webster].

Usually, an image is intended to be a representation of objects or

scenes on a bidimensional support
I painting,
I picture,
I graph.

The concept can be extended to a volumetric representation (e.g.,

a statue, but also the inside).

Definition of digital image

I In Mathematics, an image is defined as a function,

f : Rm → Rn .
I Usually, m = 2 and, in the simplest case, n = 1.
I When the spatial coordinates and the function value are finite
and discrete, the image is called digital.

f : Zm → Zn

I The elements composing a digital image are called “pixels”

(contraction of picture element) or “pels” (or even “image
elements”).

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 2

.

What is image processing?

I Roughly speaking, image can be considered what is perceived

through the sight.
I However, “seeing” is a complex task, and it results from a
very variegated processing chain realized by the brain.
I In the same way, the image processing is a discipline with a
fuzzy contour.
I Besides, it makes use of and provides techniques to other
disciplines.
I At least three disciplines are overlapping:
I image processing
I image understanding
I computer vision

What is image processing? (2)

I Generally, three processing categories, which are characterized

by their abstraction level, are acknowledged:
I low level (image enhancement);
I medium level (feature extraction);
I high level (object recognition).

I Example: OCR (optical character recognition)

I low level: noise suppression, binarization;
I medium level: character segmentation;
I high level: character and text structure recognition.

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 3

.

Which application for image processing?

I In several application fields, information is acquired as images.

I Image processing is used for:
I making data interpretation easier;
I extracting information.
I Some application fields:
I medicine;
I industrial:
I quality control;
I automation;
I security:
I biometric application;
I license-plate recognition;
I topographic survey;
I astronomy;
I . . . and many more!

Which application for image processing? (2)

I For example, in astronomical observation.

Harvard College Observatory

Computers, 1890s.
Stars cataloging were realized with
manual methods and were based on
photographic plates.
https://en.wikipedia.org/wiki/Harvard_Computers

I A lot of tasks are not easily automated:

Galaxy Zoo Project.

Volunteers (with no special
astronomy knowledge) are
required for galaxy classification.
http://www.galaxyzoo.org/

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 4

.

Which images?

I A possible classification can be based on the radiating source:

I electromagnetic waves;
I acoustic waves;
I ultrasound;
I electron beam.

Electromagnetic spectrum

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 5

.

Gamma-ray imaging

a b
c d

(a) gamma-ray imaging

by radioactive isotope
injection;
(b) positron emission
tomography (PET);
(c) Gamma radiation
emitted by a nuclear
reactor valve;
(d) Cygnus Loop.

NB: (c) and (d) captu-

re the natural radiation of
the observed object.

X-ray imaging

a
d
b
c e

(a) radiography;
(b) angiography;
(c) computerized axial to-
mography (CAT);
(d) X-ray image of an
printed circuit board for
inspection during produc-
tion;
(e) Cygnus Loop.

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 6

.

Ultraviolet imaging

a b
c

(a) normal corn;

(b) smut corn;
(c) Cygnus Loop.

In UV imaging, the fluo-

rescence phenomenon is
often exploited.

Visible light imaging

a b c
d e f

Microscopy
(a) Taxol (anticancer
agent);
(b) cholesterol;
(c) microprocessor;
(d) nickel oxide thin
film;
(e) surface of an audio
CD;
(f) organic supercon-
ductor.

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 7

.

Visible light imaging (2)

a b
c d
e f

Industrial automation
(a) missing components;
(b) missing pills;
(c) right quantity;
(d) defects detection (too
many bubbles);
(e) defects detection (so-
me burnt flake?);
(f) defects detection (lens
with distortion).

Visible light imaging (3)

a b
c
d

Security and traceability

(a) fingerprint;
(b) banknote counting
and serial number trac-
king;
(c) Automatic license pla-
te recognition;
(d) Automatic license pla-
te recognition (with per-
spective correction).

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 8

.

Visible light and infrared imaging

a b c
(a) blue; (b) green; (c) red;
d e f g (d) near IR; (e) short-wave IR;
(f) thermal IR; (g) mid-wave IR.

Visible light and infrared imaging (2)

The weather sa-

tellite images
are taken from
several bands.
This LANDSAT
image pictures the
Katrina hurricane
(2005) in the
visible and infrared
bands.

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 9

.

IR satellite imaging

The infrared band emission provides clues of human activities.

Radar satellite imaging

The microwaves can pass through the clouds and some layers of
plants, ice, and sand. A space-borne radar emits a pulse and,
through an antenna, detects the reflected wave.

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 10

.

Radio wave imaging

Radio waves are used in medicine, for instance, for magnetic

resonance imaging (MRI).

Multi-spectral imaging

The Crab pulsar in several sub-bands of the spectrum. The

same object can appear very different, since its emitted
electromagnetic radiation is not uniform in the spectrum.

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 11

.

Acoustic wave imaging

Detecting the echo from acoustic waves, the reflective properties

of the ground (or seabed) can be traced. This allows to identify
the stratified materials in the ground.

Ultrasound imaging

a b Ultrasound waves are used in medicine for echography.

c d (a) e (b) fetus; (c) thyroids; (d) muscle layers showing
lesion.

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 12

.

Electrons beam imaging

a b

Electrons beams are used in two microscope techniques:

I transmission electron microscopy (TEM), where the
fraction of the beam that pass through the object is
acquired;
I scanning electron microscope (SEM), where the interaction
between the object and the electrons beam is detected.
(a) tungsten filament burnt; (b) heat damaged integrated circuit
(the white fibers results from the thermal destruction).

Other imaging techniques

Through image processing techniques, new kind of images can be

obtained. For example, the depth images (aka range images, 2 12 D
images).

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 13

.

Example: KinectTM

Futuristic (?)

Mind reading? Processing fMRI images.

I fMRI recording of subject while watching movie clips

I model of the brain fed with the recording
I model output as images

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 14

.

Homeworks and suggested readings

DIP, Sections 1.1, 1.2, 1.3

I pp. 1–24

This is your brain on fMRI

I http://spectrum.ieee.org/geek-life/
tools-toys/
this-is-your-brain-on-fmri
I http://dx.doi.org/10.1109/MSPEC.
2012.6189569

Stefano Ferrari— Methods for Image processing— a.y. 2022/23 15