Istituto di Fisica Applicata “Nello Carrara”
National Research Council (CNR)
Firenze, Italy -
www.ifac.cnr.it
Image quality issues in
Image quality issues in
digitization
digitization
projects of historical documents
projects of historical documents
Franco Lotti
Franco Lotti
IFAC activity in Italian digitisation projects of ancient docume
IFAC activity in Italian digitisation projects of ancient docume
nts
nts
¾
IMAGO-II
projects
(1998-2000).
State Archives of Florence and Lucca
(more than 100,000 parchment rolls of the
Diplomatico
fonds, VIII-XIV
century).
http://www.archiviodistato.firenze.it/progetti/attivite.htm
http://www.comune.lucca.it/archiviostato/supp-inf.html
¾
Candido
project
(2000-2003).
University Library of Pisa
(about 85.000
images of manuscripts, drawings, printed books on paper and parchment,
collection of letters of famous personalities, periodicals from XIV to XIX
century).
http://www.cab.unipd.it/eventi/pisa.php3
;
http://www.pisa.sbn.it
¾
Datini
project
(2001-2004).
State Archive of Prato
(450,000 images of
Francesco Datini’s collection of private and trade letters, bookkeeping notes,
management books, etc. – XIV century).
http://www.archiviodistato.prato.it
¾
Monumenti Nazionali
project
(2002-2004).
State Archive of Frosinone
(feasibility study for the digitisation and web consultation of parchments
preserved by various Middle Ages abbeys and monasteries in Central Italy).
D
D
igitisation
igitisation
p
p
olicies
olicies
of the European Union
of the European Union
Lund
Lund
-
-
4th April 2001: EC expert’s meeting
4th April 2001: EC expert’s meeting
to accomplish coordination mechanisms of digitisation
to accomplish coordination mechanisms of digitisation
policies and programmes across Europe
policies and programmes across Europe
⇒
Lund Principles
Lund Principles
⇒
Lund Action Plan
Lund Action Plan
“The European culture can be freely
accessible through the digitisation of
cultural content.
This will also support and promote the
cultural difference in a global scenario”
D
D
igitisation
igitisation
p
p
olicies
olicies
of the European Union
of the European Union
MINERVA PROJECT
MINERVA PROJECT
(coordinator
(coordinator
:
:
Italy)
Italy)
a network of Member States’ Ministries
a network of Member States’ Ministries
--
to implement the Lund Action Plan;
to implement the Lund Action Plan;
--
to discuss, correlate and harmonise activities in
to discuss, correlate and harmonise activities in
digitisation of cultural and scientific content;
digitisation of cultural and scientific content;
--
for creating agreed European common
for creating agreed European common
recommendations and guidance about:
recommendations and guidance about:
--
digitisation
digitisation
--
metadata
metadata
--
long
long
-
-
term accessibility
term accessibility
Good practices in digitisation projects
Good practices in digitisation projects
- Analysis of d
ocument
ocument
s
s
--
Safe document treatment and management
Safe document treatment and management
--
Definition of user’s requirements
Definition of user’s requirements
THE FIRST STEPS
THE FIRST STEPS
--
Q
Q
uality
uality
of acquired images
of acquired images
--
Efficient and fast retrieval
Efficient and fast retrieval
Simplified layout of a
Simplified layout of a
digitisation
digitisation
project
project
IMAGE ACQUISITION METADATA COMPRESSION . . . .
Docu-ment
Master
Master
archive
archive
WEB
WEB
i
i
mages
mages
DB
DB
LAN
LAN
i
i
mages
mages
DB
DB
Index DB
Index DB
A
A
B
B
Indexing
Indexing
LAN
LAN
server
server
I
I
NTRANET
NTRANET
+
• Acquisition methodology and hardware
requirements
• Ongoing image quality assessment and tests
• Indexing - formatting and organisation of metadata
• Accessibility and dissemination strategy
• Maintenance plan
Development of the project
Development of the project
COSTS
Image quality
Image quality
⇒
Image size
⇒
Storage
Good practices in digitisation projects
Good practices in digitisation projects
⇓
⇓
--
Illumination system
Illumination system
--
Type of benches and document
Type of benches and document
supports
supports
--
Sensor characteristics
Sensor characteristics
--
Preprocessing
Preprocessing
--
Calibration procedures and stability
Calibration procedures and stability
--
Optical and mechanical performances
Optical and mechanical performances
--
Compression
Compression
F
F
actors affecting image quality
actors affecting image quality
--
Material typology
Material typology
--
Binding type
Binding type
--
Status of conservation and fragility
Status of conservation and fragility
DOCUMENT CHARACTERISTICS:
DOCUMENT CHARACTERISTICS:
--
Light sensitivity
Light sensitivity
--
Size and type of content
Size and type of content
--
B&W, grey or colour
B&W, grey or colour
Choosing the right instrumentation
Choosing the right instrumentation
--
S
S
ampling
ampling
rate (
rate (
ppi
ppi
-
-
pixel per inch)
pixel per inch)
--
Number of elements and geometry (
Number of elements and geometry (
li
li
near,
near,
matrix)
matrix)
--
Spatial resolution, MTF
Spatial resolution, MTF
(cycles/mm)
(cycles/mm)
--
Bit depth (
Bit depth (
bpp
bpp
-
-
bit per pixel)
bit per pixel)
SENSOR CHARACTERISTICS:
SENSOR CHARACTERISTICS:
--
Type of scan (one shot, three s
Type of scan (one shot, three s
hots
hots
)
)
--
Type of view (fixed
Type of view (fixed
-
-
planetary, flat
planetary, flat
-
-
bed, …)
bed, …)
Choosing the right instrumentation
Choosing the right instrumentation
Sine
Sine
-
-
wave patterns:
wave patterns:
to evaluate the
to evaluate the
modulation transfer
modulation transfer
function (MTF)
function (MTF)
Square
Square
-
-
wave
wave
patterns: to
patterns: to
evaluate the
evaluate the
contrast
contrast
transfer
transfer
function (CTF)
function (CTF)
Test charts for spatial resolution
Test charts for spatial resolution
Evaluation of the sensor performances
Evaluation of the sensor performances
Evaluation of the sensor performances
Evaluation of the sensor performances
Computing the Modulation
Computing the Modulation
Transfer Function (MTF), by the
Transfer Function (MTF), by the
acquisition of sine
acquisition of sine
-
-
wave test
wave test
charts, is the correct way to
charts, is the correct way to
evaluate the
evaluate the
spatial resolution
spatial resolution
.
.
The evaluation of the Contrast
The evaluation of the Contrast
Modulation Function (CTF)
Modulation Function (CTF)
,
,
by the
by the
acquisition of square
acquisition of square
-
-
wave test
wave test
charts,
charts,
tends to over
tends to over
-
-
estimate the
estimate the
quality of the sensor.
quality of the sensor.
MTF and CTF
MTF and CTF
MTF
CTF
Colour
Colour
Test charts
Test charts
Evaluation of the sensor performances
Evaluation of the sensor performances
Choosing the right instrumentation
Choosing the right instrumentation
Light, UV fraction and temperature measurements
Light, UV fraction and temperature measurements
Image compression
Image compression
A
A –
–
Reversible (
Reversible (
Lossless
Lossless
) methods
) methods
:
:
•
•
Bitmap
Bitmap
•
•
TIFF uncompressed
TIFF uncompressed
•
•
TIFF LZW (Lempel Ziv Welch)
TIFF LZW (Lempel Ziv Welch)
•
•
PNG
PNG
•
•
GIF
GIF
•
Image compression
Image compression
B
B –
–
Lossy
Lossy
methods
methods
:
:
•
•
JPEG
JPEG
•
JPEG Compression 1: 20
JPEG Compression 1: 20
-
-
artifacts
artifacts
300
JPEG Compression 1: 20
JPEG Compression 1: 20
-
-
artifacts
artifacts
1 cm
JPEG Compression 1:50
JPEG 2000 file format
JPEG 2000 file format
•
•
Compatibility with ISO standard
Compatibility with ISO standard
•
•
Openness
Openness
•
•
Interoperability (systems compliant with the standard)
Interoperability (systems compliant with the standard)
•
•
Non
Non
-
-
proprietary
proprietary
•
•
Supports embedded metadata
Supports embedded metadata
•
•
IPR and XML boxes for property and vendor information
IPR and XML boxes for property and vendor information
•
•
Scalability, both in quality and resolution
Scalability, both in quality and resolution
•
•
Lossy
Lossy
and
and
lossless
lossless
decompression
decompression
•
•
Progressive display
Progressive display
•
Example:
Example:
Comparison
Comparison
between
between
JPEG
JPEG
and
and
J2
J2
K
K
Case A
Case A
:
:
Bit rate = 2.55 (CR = 9.4)
Bit rate = 2.55 (CR = 9.4)
Ggood
Ggood
quality for LAN
quality for LAN
consultation (intranet)
consultation (intranet)
Case B
Case B
:
:
Bit rate = 0.46 (CR = 51,8)
Bit rate = 0.46 (CR = 51,8)
Low quality: for web
Low quality: for web
dissemination (internet)
dissemination (internet)
Datini
Datini
Project
Project
(State Archive of Prato)
High
High
-
-
Q JPG
Q JPG
–
–
(intranet)
(intranet)
Low
Low
-
-
Q JPG
Q JPG
–
–
(internet)
(internet)
Example:
0.39
186
9.53
51.83
0.46
Low quality
J 2K
0.81
40
4.12
9.40
2.55
High Quality
J2K
0.33
187
10.93
51.83
0.46
Low quality
JPEG
0.73
55
5.05
9.40
2.55
High Quality
JPEG
Q-index
(**)
PE (*)
RMSE
(*)
CR
Bit rate
(*) Green channel
Example:
Some examples
Some examples
1 - Parchments
3 - Printed paper
4 - Manuscripts
5 - Drawing
2 - Seals
Example 1
Example 1
-
-
Project:
Project:
IMAGO II
IMAGO II
Digitisation
Digitisation
of the
of the
Diplomatico
Diplomatico
fonds
fonds
State Archive of
State Archive of
Florence:
Florence:
More than
More than
140,000
140,000
Parchment rolls
Parchment rolls
681 provenances
681 provenances
VIII
VIII
–
–
XIX century
XIX century
A box of parchment rolls of the Diplomatico.
A box of parchment rolls of the Diplomatico.
State Archive of Florence
Example 1
Example 1
-
-
Project:
Project:
IMAGO II
IMAGO II
Digitisation
Digitisation
of the
of the
Diplomatico
Diplomatico
fonds
fonds
State Archive of Lucca:
State Archive of Lucca:
About
About
23
23
,
,
000
000
parchment rolls
parchment rolls
70 provenances
70 provenances
VIII
Example 1
Example 1
-
-
Project:
Project:
IMAGO II
IMAGO II
Digitisation
Digitisation
of the
of the
Diplomatico
Diplomatico
fonds
fonds
To ensure safety of
To ensure safety of
document handling,
document handling,
suitable sliding
suitable sliding
windows have been
windows have been
designed for the
designed for the
acquisition of
acquisition of
parchment rolls on
parchment rolls on
both sides
both sides
(recto and verso)
(recto and verso)
Presentation 1:1
Presentation 1:1
Example 1
Example 1
Project:
Project:
IMAGO II
IMAGO II
Digitisation of parchment
Digitisation of parchment
manuscripts
manuscripts
Sensor :
Sensor :
Matrix, Three shots
Matrix, Three shots
(no interpolation)
(no interpolation)
Spatial sampling:
Spatial sampling:
200
200
ppi
ppi
Colour depth:
Colour depth:
36 bit/pixel,
36 bit/pixel,
rescaled to 24
rescaled to 24
bpp
bpp
Compression:
Compression:
JPEG
JPEG
average CR: ~ 10
average CR: ~ 10
Access:
Access:
Intranet.
Intranet.
The quality of compressed
The quality of compressed
images showed a hardly
images showed a hardly
appreciable degree of loss,
appreciable degree of loss,
Enlargement: 4:1
Enlargement: 4:1
Example 2
Example 2
-
-
Project:
Project:
IMAGO II
IMAGO II
Digitisation
Digitisation
of the
of the
Diplomatico
Diplomatico
fonds
fonds
ACQUISITION OF
ACQUISITION OF
WAX SEALS
WAX SEALS
Sensor :
Sensor :
Matrix, Three shots
Matrix, Three shots
(no interpolation)
(no interpolation)
S
S
p
p
a
a
tial
tial
sampling
sampling
:
:
4
4
00
00
ppi
ppi
Colour depth:
Colour depth:
36 bit/pixel
36 bit/pixel
rescaled to 24
rescaled to 24
Illumination: 2 flashes,
Illumination: 2 flashes,
a
a
symmetric
symmetric
C
C
ompression
ompression
: JPEG
: JPEG
,
,
average CR
Example 3
- CANDIDO
Project
–
University Library of Pisa
Periodicals (XIX Century)
Sensor :
Sensor :
Three
Three
-
-
linear scanner
linear scanner
S
S
patial
patial
sampling
sampling
:
:
200
200
ppi
ppi
Colour depth:
Colour depth:
24 bit/pixel
24 bit/pixel
C
C
ompression
ompression
:
:
JPEG, CR
JPEG, CR
≈
≈
10
10
Access
Example 4
- CANDIDO
Project
–
University Library of Pisa
Correspondance Rosselmini-Gualandi
Sensor
Sensor
: Scan back, Three
: Scan back, Three
-
-
linear
linear
array
array
S
S
patial
patial
sampling
sampling
:
:
300
300
ppi
ppi
Colour depth
Colour depth
: 48 bit/pixel
: 48 bit/pixel
rescaled to 24 bpp
rescaled to 24 bpp
C
C
ompression
ompression
: JPEG
: JPEG
, CR
, CR
≈
≈
10
10
Access
Example 4
- CANDIDO
Project
–
University Library of Pisa
Example 4
- CANDIDO
Project
–
University Library of Pisa
Example 4
- CANDIDO
Project
–
University Library of Pisa
Example 5
- CANDIDO
Project
–
University Library of Pisa
Rosellini - Drawings
Sensor
Sensor
: Scan back, Three
: Scan back, Three
-
-
linear
linear
array
array
S
S
patial
patial
sampling
sampling
:
:
30
30
0
0
ppi
ppi
Colour depth
Colour depth
: 48 bit/pixel,
: 48 bit/pixel,
rescaled to 24 bpp
rescaled to 24 bpp
C
C
ompression
ompression
: JPEG
: JPEG
, CR
, CR
≈
≈
10
10
Access
Conclusions
Conclusions
- Image quality depends on a number of factors related to the
capture methods and the further processing steps.
- Difficult to weight all those factors in a rigorous way.
- Suitable test procedures advisable, also tuned to the
specific targets.
Key points:
Correct acquisition of masters
Adaptive and progressive compression and smart,
robust IRP methods
Istituto di Fisica Applicata “Nello Carrara”
National Research Council (CNR)
Firenze, Italy -
www.ifac.cnr.it
… Thank you for your kind
attention !