Historical Practice
Documents -- Printed matter, manuscripts
Legible, printable reference image
Pictorial items -- Prints and photographs
Reference surrogate
Current Practice -- Documents
For-information-only class
Routine must-be-scanned-bound class
Derived-from-microfilm class
No OCR class
Legible, printable reference image
Examples:
- Journals of the U.S. Senate
- WPA typescripts
- George Washington Papers manuscripts (bulk)
- Unexceptional books 1880s and later
Current Practice -- Documents
Rationale for legible-only practice
- Legibility alone sufficient for document
- Impossible to get a perfect image
- Less likely to be selected for "art" reproduction
- Include printed halftones
- No OCR contemplated
Current Practice -- Documents
Typical master/archival image types for legible-only
Lossless compressed bitonal page images
300-400 dpi, CCITT Group IV
Lossy compressed tonal (grayscale) page images
200-300 dpi, JPEG (not fully satisfactory)
Current Practice -- Documents
Art-of-printing class
Rare, large, and/or valuable class
OCR class (no LC examples)
High resolution archival capture
Examples:
- Broadsides in Rare Book division
- Manuscript treasures (including some George Washington)
Current Practice -- Documents
Rationale for high resolution practice
High artifactual value
Hard to handle
Likely to be selected for "art" reproduction
OCR contemplated
Current Practice -- Documents
Typical archival (master) image types
for high resolution
Uncompressed tonal (grayscale or color) page images
300 (or greater) dpi, uncompressed (or LZW?)
Lossless compressed bitonal page images
600 dpi, CCITT Group IV
Mantra
For viewing and reference-use printing,
high quality digital images are better than microfilm images.
For fine reproduction, e.g., printed publications, high quality
digital images
can permit the production of printing plates as successfully as copy
photographs.
Once systems demonstrate capability to migrate data, high quality digital
images will be accepted for long-term preservation.
Current Practice -- Pictorial
When the original item is to be scanned
High resolution archival capture
Example:
HABS/HAER architecture negatives
Civil War prints
Current Practice -- Pictorial
When copy intermediate is to be scanned
Moderate resolution capture
Example:
Copy negative file
Current Practice -- Pictorial
Archival (master) image
Higher resolution
5000 pixel (long side) tonal image (grayscale or color), uncompressed
Moderate resolution
3000 pixel (long side) tonal image (grayscale or color), uncompressed
(Yes, these are on the high side.)
Mantra
For fine reproduction, e.g., printed publications, high quality digital images
can permit the production of printing plates as successfully as copy
photographs.
High quality digital images can fill all the user needs that are presently
provided by photographic copy negatives or interpositives.
Once systems demonstrate capability to migrate data, high quality digital
images will be accepted for long-term preservation.
Document Resolution Measurement Units
LC tends to describe document or page images in terms of the
dimensions of the original
Examples: 150 dpi, 300 dpi, 600 dpi
Document Resolution Measurement Units
LC typical bitonal images:
8.5x11 page/300 dpi 2550x3300
6x9 page/300 dpi 1800x2700
Cornell preservation bitonal images:
8.5x11 page/600 dpi 5100x6600
6x9 page/600 dpi 3600x5400
Pictorial Resolution Measurement Units
LC tends to describe pictorial images in terms of the "dot count"
without regard to the dimensions of the original.
Archival files in next contract:
5UA 5000x4000 window
4UA 4000x3000 window
3UA 3000x2000 window
Pictorial Resolution Measurement Units
Reference files in next contract:
CRI 640x480 window
Videodisc-era carryovers
560x420 window
Thumbnails
150x150 window
For Comparison
VGA 640x480
LC W95 desktop 800x600
SVGA 1024x768
Need for Multiple Versions
Archival ("proto-preservation")
Reference and/or printing
Thumbnail/inline
On-the-fly generation begins to emerge
Multiple Versions -- Documents
Legible-only class
Page-turning device ("binder")
Inline GIF tonal image
400-800 pixels across
Master/archival image
300 dpi bitonal
200-300 dpi tonal compressed
Multiple Versions -- Documents
Broadside worst-case oversize (under discussion, no decision yet)
Bib record thumbnail (first page only)
GIF tonal, 150 pixels across
Searchable text inline
GIF tonal, 250 pixels across
Page-turning device ("binder") inline
GIF tonal, 400-800 pixels across
Reference image (fetchable)
JPEG tonal (some examples 2500 pixels across)
Master/archival image
Uncompressed 300 dpi tonal (some examples 5000 pixels across)
Multiple Versions -- Pictorial
Pictorial (highest quality type)
Bib record thumbnail
GIF tonal, 150 pixels across
filename ends "t"
Reference image (fetchable)
JPEG tonal, 640x480 window
filename ends "r"
Master/archival image
Uncomp (LZW?) 5000 pixels across
filename ends "u"
Compressed display image (under discussion)
JPEG 5000 pixels across
filename ends "v"
Producing Derivatives
Bitonal Document Master Images
Produce inline tonal image (interlaced GIF)
Gray (4 bits deep)
Blur
Rescale
Sharpen
Most promising for on-the-fly derivation
Producing Derivatives
Tonal Document Archival Images
Produce inline tonal image (interlaced GIF)
Drop grayscale to 4 bits deep
Color drop to 8 bits deep (indexed color)
Rescale
Sharpen (?)
Less promising for on -the-fly
Producing Derivatives
Pictorial or high quality document
Master/archival
no non-reversible enhancements, e.g., no sharpening
greater bit depth (?)
Reference image
rescale
sharpen
may contrast stretch
may use non-linear tonal distribution for monitor display
Thumbnails
drop to 8 bits deep
creative use of blur and sharpen
Capture Devices
Flatbed scanners
Practical when handling or conservation permits
May go to 11x17 inches (map scanner to 32x24 inches)
High end devices have greater ability with broader brightness range include
control of black and white points
Capture Devices
Overhead devices
Some are camera mounted linear CCD arrays
Some are camera mounted area-capture CCD arrays
Preferred when items cannot lie flat or cannot be inverted
Typically have excellent software
Capture Devices
The greater the power and control, the more skill required.
View camera controls
Software setup
Management of ambient light
More complex workstation
Increased investment for high end devices means
more headaches re: obsolescence.
Bonus Slides
Gleanings from Pictorial RFP
Image Tonal Range -- 1
The tonal range of the delivered digital images delivered shall be
representative of the original scene or artifact or, in the case of images whose
source is a photographic negative, of the expected representation of the
original scene when the negative is reproduced as a positive print.
For prints (black-and-white or color), the objective is to reproduce the items as
they exist in the collection.
For negatives and original positive transparencies , the objective is to create a
positive image in a manner that may be compared to creating a print
(black-and-white or color) in a darkroom.
Bonus Slides
Gleanings from Pictorial RFP
Image Tonal Range -- 2
Providing images with acceptable tonal qualities shall require that the
scanning operator exercise judgement when producing the images. The
operator's judgement required to achieve the required outcome shall be
especially critical when imaging color items. Utilization of general imaging
industry standards and those as agreed to and established during the contract
startup and testing phase shall be followed. Additionally, consultation with
Library staff may, when necessary, be required in order to ensure that
appropriate operator judgements are made throughout the run of a particular
batch or scan group.
Bonus Slides
Gleanings from Pictorial RFP
Tonal Value
For a black-and-white scene with a typical range of brightness, e.g., a
landscape in daylight or a conventional studio portrait, a histogram shall show
continuity of sampling and shall include values ranging from black to white.
Some pixel values shall fall in the range of 5-12 (black) while some values shall
fall in the range 243-250 (white).
For color images with typical scene brightness that include white or black
elements, similar values shall be provided for RGB (red, green, blue)
renderings of white or black, i.e., a white area shall have values of r=243-250,
b=243-250, and g=243-250 and a black area shall have values of r=5-12,
b=5-12, and g=5-12.
Bonus Slides
Gleanings from Pictorial RFP
Cropping -- 1
The Library wishes to provide researchers with a reproduction of the entire
original item. Thus, images shall be framed and cropped to show the entire
original item and beyond the item's edges. For negatives or other transmitted
light items, each digital image shall reproduce that item's actual-image area,
the border on the film that surrounds the image area, and a portion of the
background (light box or scanner top) beyond the edge of the film. A similar
approach shall be followed for reflected-light items; the whole print, whole
mount, and a portion of the background (beyond the mount) shall be
reproduced.
In the delivered images, the amount shown beyond the edge of the item shall be
no less than 1.5 percent of the dimension of the long side image. Thus, for a
3,000 x 2,000-pixel image, the border beyond the reproduction of the original
item shall consist of approximately 35 pixels on all four sides; for a 640 x
480-pixel image, the border shall consist of approximately 10 pixels on all four
sides.
Bonus Slides
Gleanings from Pictorial RFP
Cropping --2
Multiple Items on a Single Mount
When multiple prints are mounted on a single board, e.g., eight Civil War
portraits on one cardboard mount or three small prints mounted on a single
scrapbook page, each image on a multiple-print mount shall be captured
separately. Framing for each print shall extend beyond the image proper.
Specific task orders may additionally require the production of an image that
captures the full physical item, i.e., a Civil War mount with eight portraits or a
full scrapbook page with three prints.