From Analog to Digital

1
From Analog to Digital
2
What are Digital Images?

• Electronic snapshots taken of a scene or scanned
  from documents
• samples and mapped as a grid of dots or picture
  elements (pixels)
• pixel assigned a tonal value (black, white,
  grays, colors), represented in binary code
• code stored or reduced (compressed)
• read and interpreted to create analog version

3
CUL Bias on Image Capture

• Create rich images that are useful over time in
  the most cost-effective manner.
• Set conversion requirements greater than
  immediate application
• Promote reuse of content
• Enable sharing of comparable and trusted
  resources across disciplines, users, and
  institutions

4
Why Rich Digital Masters?

• Preservation
• Original may only withstand one scan
• Maintenance of digital files
• Cost
• One scan may be all that is affordable
• Conversion costs dwarfed by other costs
• Access
• Many from one
• The richer the file, the better the derivative in
  terms of quality and processibility

5
How to determine whats good enough?

• Connoisseurship of document attributes
• Identify key information content
• Objectively characterize or measure attributes
  size, detail, tone, and color
• Appreciate imaging factors affecting quality and
  cost
• Translate between analog and digital
• Equate measurements to digital equivalencies and
  corresponding metrics, e.g., detail size ?
  resolution ? MTF

6
CULs Approach to Imaging No More, No Less
desired point of capture
image quality and utility
Image requirements and cost
7
Digital Image Quality is Governed By

• resolution and threshold
• bit depth
• color management
• image enhancement
• compression and file format
• system performance

8
Resolution

• Determined by number of pixels used to represent
  the image
• Increasing resolution increases level of detail
  captured and geometrically increases file size

zoom in
9
Effects of Resolution
600 dpi 300 dpi 200 dpi
10
Threshold Setting in Bitonal Scanning

• defines the point on a scale from 0 to 255 at
  which gray values will be interpreted either as
  black or white

11
Effects of Threshold

threshold 60
threshold 100
12
Bit Depth

• Determined by the number of binary digits (bits)
  used to represent each pixel

8-bit
24-bit
1-bit
13
Bit Depth

• increasing bit depth increases the level of gray
  or color information that can be represented and
  arithmetically increases file size
• Bit depth, dynamic range, and color appearance

14
Utilizing Sufficient Bit-Depth
3-bit gray
8-bit gray
15
Utilizing Sufficient Bit Depth
8-bit color
24-bit color
16
Bit Depth vs. Dynamic Range

• The range of tonal difference between lightest
  light and the darkest dark

17
Mapping Tones Correctly Use of Histograms
18
Representing Color Appearance
Color Shift Towards Red
Balanced Color
19
Image Enhancement

• Image editing to modify or improve an image
• filters (brightness, contrast, sharpness, blur)
• tone and color correction
• Use raises concerns about fidelity and
  authenticity

20
Effects of Filters
no filters used maximum enhancement
21
Image Editing
22
Compression

• reduces file size for processing, storage,
  transmission, and display
• image quality may be affected by the compression
  techniques used and the level of compression
  applied

23
Compression Variables

• lossless versus lossy compression
• proprietary vs. open schemes
• level of industry support
• bitonal vs. gray/color
• see attributes of common compression techniques
  at www.library.cornell.edu/preservation/tutorial/
  presentation/table7-3.html

24
Effects of JPEG Compression
300 dpi, 8-bit grayscale uncompressed TIFF
JPEG 18.51 compression
25
Compression /File Format Comparison
GIF (lossless) File Size 60 KB
JPEG (lossy) File Size 49 KB
images courtesy of Edison Papers
26
File Formats

• Consist of both the bits that comprise image
  information and header information on how to read
  and interpret the file
• Image quality affected by format support for
• Bit depth
• Compression techniques
• Color management
• Hardware,software, and network support
• See common file formats chart at
• www.library.cornell.edu/preservation/tutorial/pres
  entation/table7-1.html

27
Equipment used and its performance over time

• scanners with same stated functionality can
  produce different results
• Factors affecting image quality
• Optical, mechanical, and sensing components
• Calibration
• Age of equipment
• Environment

28
Variations in Image Quality due to Scanner
Performance
300 dpi, scanner A
300 dpi, scanner B
29
Correlating Document Attributes to Image
Requirements

• Example Determining Resolution Requirements
• Whats you finest feature?
• Whats your quality requirement?
• Whats your imaging approach?

30
Case Study Brittle Books

• Variables feature size, quality, imaging
  approach
• fixed metric smallest lower case letter
• QI values 8(excellent), 5 (good), 3.6 (marginal)
• Resolution key to text capture, e.g., dpi
• Bitonal QI formula for text
• DPI 3QI/.039h
• 600 dpi 1-bit capture adequately preserves
  informational content of cleanly produced text
  and supports image processing (e.g., OCR)

31
(No Transcript)
32
(No Transcript)
33
Textual Documents May Require Tonal Capture

• Pages badly stained
• Pages exhibit low contrast between text and
  background
• Fine features not fully resolved
• Pages contain complex graphics, color or
  important contextual information
• Gray/color QI formula for text
• Dpi 2QI/.039h

34
Determining Resolution Requirements beyond Text

• Stroke width
• Finest scale
• Visual perception

35
Defining Detail as Stroke

• Edge-based representations
• Variables feature size, quality, imaging
  approach
• fixed metric width of finest line, stroke, dot,
  or marking
• QI values based on sampling frequency
  2(excellent), 1.5 (good), 1(marginal)
• Resolution and bit depth key to quality

36
Defining Detail as Stroke

• QI formulas for stroke
• Gray/color dpi QI/.039w
• QI dpi x .039w
• Bitonal dpi (1.5QI)/.039w

37
Adequately Rendered Stroke
38
Inadequately Rendered Stroke
39
Example

• Manuscript page with finest significant stroke
  measuring .2mm, which must be fully captured
• Gray/color dpi QI/.039w
• QI2, w.2
• Dpi 2/.039(.2) 256 dpi

40
Defining Detail as Scale

• Smallest significant scale or repeatable pattern,
  e.g., knots in a rug vs. strands in the thread
• Can result in very high resolution requirements
  (e.g., photographs, book illustrations)
• Halftones scan at 4 times the screen ruling,
  utilize special descreen/rescreening, or scan in
  grayscale (400 dpi is a good default)

41
Halftone Scanned at 150 DPI
42
Halftone Scanned at 400 DPI
43
Defining Detail Based on Visual Perception

• Human eye can detect details approximately 1/215
  inch wide
• Finer details are optically averaged
• Using two pixel rule, visual perception
  requirements met at 430 dpi
• Illustrated Book Study 400 dpi grayscale
  recommendation

44
www.library.cornell.edu/preservation/illbk/AdComm.
htm y.
45
Detail Represented at 400 dpi
46
Translating Between Digital Resolution and
Scanner Performance

• Detail capture governed by resolution, threshold,
  bit-depth, and system performance
• Sampling resolution (DPI) is not a true indicator
  of image quality, although it may suffice for
  scanning in the 300-600 dpi range
• Monitor Emerging System Performance Metrics
  (e.g., MTF)

47
Aligning Document Attributes with Digital
Requirements

• Identify key document attributes
• Tone, color, and detail
• Characterize them, if possible through objective
  measurements
• Determine quality requirements and tolerance
  levels
• Translate between analog and digital and between
  scanning requirements and scanning performance

48
Aligning Document Attributes with Digital
Requirements

• Calibrate scanner with targets and software
• Calibrate the rest of the system
• Control lighting and environment
• Scan appropriate targets with documents
• Evaluate images against originals

49
Aligning Document Attributes with Digital
Requirements

• Minimize post-processing in the master image
• Save in TIFF avoid lossy compression
• Maintain scanning metadata
• Monitor emerging image quality metrics

50
One Size Does Not Fit All!

• Different document types will require different
  scanning equipment and processes
• The more complex the document, the higher the
  conversion/access requirements
• Scan the original whenever possible
• No standards for image conversion guidance
  rather than guidelines
• Notion of long-term utility and
  cross-institutional resources gaining ground