Posters on demand at the National Gallery

1
Posters on demand at theNational Gallery

• John Cupitt
• Scientific Department
• The National Gallery
• London
• HP Art Science 2003, Florence
• http//www.vips.ecs.soton.ac.uk
• http//cima.ng-london.org.uk/john/POD

2
Overview

• What is POD
• Digitisation
• Poster image production
• What is a colour match
• Next

3
What is POD

• Kiosk in National Gallery Sainsbury Wing shop,
  public opening Wednesday 18 June
• Visitors select painting and a print size poster
  printed in about 10 minutes
• Choice of 1,200 paintings
• 10 for A4, 15 for A3, 25 for A2
• Specials 50 150
• Project funded by Hewlett-Packard and The
  National Gallery

4
(No Transcript)
5
Technical overview

• Productivity semi-gloss paper, UV inks, 42 HP
  DesignJet 5500ps
• sRGB JPEG images, PDF layout
• Driverless printing with canned profile
• Content generated automatically only 3 paintings
  proofed
• Posters are colour facsimilies (under the right
  viewing conditions)
• Posters that look like paintings

6
The MARC camera

• Studio digital camera built during the EU-funded
  MARC project, 92 95
• Version 2 built 98 2000
• Takes a 10,000 by 10,000 pixel image in 2m 40s
• 12 bit, D65 CIELAB colour, 3?E with HMI
• Version 3 commercialised by Sinar

7
Scanning campaign

• Collection scanned 2000 02 (2,500 paintings)
• Special studio built
• Two extra photographers, three extra picture
  handlers
• Archived on CD-R at three sites
• VIPS format for master image, TIFF and JPEG for
  derivative images
• Images held on-line on 2TB server

8
(No Transcript)
9
POD content generation
Staging database
Kiosk
10
Image render
Master image on server D65 CIELAB Absolute
colorimetry
Crop, rotate, resample to 200 dpi
Brighten to relative colorimetry
To XYZ
Shift white point
Convert to D50 3x3 matrix
Black down
To LAB
Usharp L
Output as Q90 sRGB JPEG
11
Colour matching

• What is a colour match? Many factors
• Viewing conditions (domestic, shop, gallery)
• Surface texture
• Consistency with other products
• Expectations
• ... and colour
• UV ink colour inconstancy

12
(No Transcript)
13
Colour accuracy

• 3?E error in camera (max 8)
• 0.7?E D65 to D50
• 3.4?E (max 5) error on print
• 4?E (max 10) end-to-end
• But we adjust the whitepoint and pull blacks down
  to improve perceived match under target
  conditions
• On the other hand, we do have a single transform
  for all posters

14
End-to-end colour errors
2 degree observer, SCE, UV 0, Minolta 2600d
15
Future work

• Expand to entire collection
• Kiosks in other parts of Gallery (and on web)
• Range of media
• Customer crops
• Remote printing
• Spectral printing or dye inks
• 3D printing
• http//vips.ecs.soton.ac.uk
• http//cima.ng-london.org.uk/john/POD