Quality Levels of Reproduction

1
Quality Levels of Reproduction

• Adolf Knoll
• National Library of the Czech Republic

2
Analogue world
Original
continuous
discrete
Archival image
Access image
Digital world
Facsimile printing or display
Preprocessed set of various quality level images
Dynamically postprocessed Image on demand

• Transition from analogue to digital first loss
  of information
• Acceptable technical parameters (resolution,
  colour depth, lossy compression)
• Acceptable colour fidelity (ICC profiles,
  calibration, metadata for still images)

3
What is necessary and what not?

• Visual representation of documents
• To deliver only the necessary quantity of
  information from the whole source
• Where is the limit between necessary and
  unnecessary?
• It is given by the purpose of communication
  facsimile printing, understanding of text, study
  of illuminations, general idea what the document
  is about, structure and character of objects on
  the digitized page, ?

4
Typical documents

• Illuminated manuscripts
• Textual manuscripts and old printed books
• Old periodicals
• Typewritten documents
• Modern journals (with colour photographs)
• Maps
• Administrative documents and forms
• Computer generated documents
• ...

5
Access images
Colour depth
Resolution
Acceptable illusion of reality
Compression
Colour fidelity
6
300 dpi 15,173 KB
Colour depth
7
256 shades of gray 5,062 KB
256 colours same size
Colour depth
8
16 shades of gray 2,531 KB
Colour depth
9
Black-and-white 2 colours 637 KB
Colour depth
10
80 dpi 3.75 times smaller file
300 dpi
Resolution
11
Possible solutions

• Decreased colour depth
• 16.7 mio colours
• 256 colours or shades of gray (1/3 of volume)
• 16 colours/shades of gray (1/6)
• 2 colours (black-and-white images, 1/24)
• Decreased resolution (down to the limit when
  necessary details are still seen)
• Combined colour depth with resolution
• Loss of information (colour or detail)

12
We have lost already a lot going from analogue to
digital
Compression

• Lossless
• Colour
• LZW 11,441 KB
• PNG 8,849 KB
• JP2 6,647 KB
• BW
• Fax Group 3
• Fax Group 4 272 KB
• Fax Group 4 166 KB
• (JBIG, JBIG2)
• JB2 (in DjVu)

• Lossy
• Colour
• DCT JPEG 318 KB
• Wavelet better, but mixed photos with text do
  not give too much space for improvement
• BW
• JBIG2-based
• JB2 98 KB
• JB2 32 KB

BEST LOSSY METHODS COMBINED Mixed Raster
Content DjVu 55 KB
13
Identical Compression Ratio
Wavelet JP2
DCT JPEG
Artefacts problem
14
MRC Mixed Raster Content
15
How about other originals?

• Typewritten text (93 dpi)
• 2,326 KB
• BW
• GIF 23 KB
• Fax Gr. 3 20 KB
• PNG 17 KB
• Fax Gr. 4 14 KB
• JB2 DjVu 8 KB
• JPEG 546 KB
• DjVu 10 KB

• Illuminated manuscript
• 225 dpi 54,654 KB
• TIFF/LZW 36,355
• PNG 28,404
• JP2 lossless 22,595
• Wavelet 1,322 (JP2, IW44 in DjVu)

16
JP2
IW44 in DjVu
1,322 KB for the entire image
DCT in JPEG
17
Colour fidelity

• ICC profile included
• Calibration table
• Printed on permanent paper with permanent colours
  
• Scanned with original
• Stored with the digital copy
• Metadata description of the image stored in a
  special SGML/XML file

18
Solutions for archiving

• Rely on well established ISO standards
• uncompressed TIFF
• TIFF/Fax Gr. 4 for black-and-white images
• JPEG
• Can rely on other well established formats
• GIF (simple graphics and bw)
• PNG (everything for efficient lossless
  compression)
• PDF for computer generated documents

19
Solutions for delivery

• The archival formats and pre-processed sets of
  images
• or
• More efficient modern solutions
• DjVu (for almost everything)
• MrSID (maps)
• Viewing comfort at the users side must be solved
  (integration of plug-ins or ActiveX components
  into web browsers)
• IE has problems with larger JPEG, but DjVu and
  MrSID very good viewing tools
• Netscape Communicator everything possible, but
  obsolete in handling
• Netscape 6.xx good plug-in and file association
  properties, but latest viewing components only as
  ActiveX (DjVu)

20
Critical points in modern solutions

• Slow encoding and high requirements for
  computation power at larger images
• Many of them depend on expensive software that
  must be purchased
• Viewing problems and integration in web browsers
• Good solutions have free browser plug-ins
  available and free encoding software for home use
  (smaller images or less functionality) they
  enable flexible display of very large files
  (DjVu, MrSID)

21
Quality Levels of Reproduction

• Usage-driven
• Creators
• Must have tools for easy and good-quality
  production
• Must provide tools for users to enable
  comfortable access (digital library solutions)
• Must preserve what they have created
• we are responsible for what we do