Attachment C - 2023 FADGI Technical Guidelines.pdf

May 2023

Technical Guidelines for Digitizing Cultural Heritage Materials Third Edition ii

Document Information Title

Editors

Technical Guidelines for Digitizing Cultural Heritage Materials: Third Edition

Thomas Rieger, Kristin A. Phelps, Hana Beckerle, Tanya Brown, Rachel Frederick, Sarah Mitrani, Patrick Breen, and Matthew Breitbart – Library of Congress Don Williams and Roger Triplett – Image Science Associates Michael Horsley – U.S. National Archives and Records Administration

Document Type Technical Guidelines

Publication Date May 2023

Source Documents Title Editor

Technical Guidelines for Digitizing Cultural Heritage Materials:

Creation of Raster Image Files September 2016 https://digitizationguidelines.gov/guidelines/FADGI%20Federal%20%2 0Agencies%20Digital%20Guidelines%20Initiative- 2016%20Final_rev1.pdf

Thomas Rieger

Document Type Technical Guidelines

Publication Date September 2016

Title Editors

Technical Guidelines for Digitizing Cultural Heritage Materials:

Creation of Raster Image Master Files http://www.digitizationguidelines.gov/guidelines/FADGI_Still_Image- Tech_Guidelines_2010-08-24.pdf

Don Williams and Michael Stelmach

Publication Date August 2010 https://digitizationguidelines.gov/guidelines/FADGI%20Federal%20%20Agencies%20Digital%20Guidelines%20Initiative-2016%20Final_rev1.pdf https://digitizationguidelines.gov/guidelines/FADGI%20Federal%20%20Agencies%20Digital%20Guidelines%20Initiative-2016%20Final_rev1.pdf https://digitizationguidelines.gov/guidelines/FADGI%20Federal%20%20Agencies%20Digital%20Guidelines%20Initiative-2016%20Final_rev1.pdf http://www.digitizationguidelines.gov/guidelines/FADGI_Still_Image-Tech_Guidelines_2010-08-24.pdf http://www.digitizationguidelines.gov/guidelines/FADGI_Still_Image-Tech_Guidelines_2010-08-24.pdf iii

Title Authors

Technical Guidelines for Digitizing Archival Records for Electronic Access: Creation of Production Master Files – Raster Images http://www.archives.gov/preservation/technical/guidelines.pdf

Steven Puglia, Jeffrey Reed, and Erin Rhodes U.S. National Archives and Records Administration

Publication Date June 2004

This work is available for worldwide use and reuse under CC0 1.0 Universal.

Acknowledgements The history of this document, as listed on the preceding page, does not begin to give credit to all of the dedicated professionals that have contributed to the body of knowledge contained in this work. It would be impossible to list all of those who have contributed over the years.

However, we wish to acknowledge organizations that have given of their time and knowledge to create this, the third formal version of the Federal Agencies Digital Guidelines Initiative (FADGI) Technical Guidelines.

First and foremost, we thank the Library of Congress for providing the institutional support required to sustain FADGI which makes this effort possible. The Library of Congress has dedicated staff support for both the Still Imaging and Audio Video Working Groups and funds significant research efforts within the Library, as well as work with expert consultants outside the Library. This document could not have been produced without the support of the Library of Congress.

The National Archives and Records Administration was the initiator of Federal efforts to establish science-based technical guidelines for cultural heritage digitization. Long before the concept of a truly digital library was understood, National Archives staff worked to establish guidance which could allow digitized materials to become the core of what we now know as digital Libraries, Archives, and Museums.

In more recent times, the role of the International Organization for Standardization (ISO) in unifying guidance has ushered in an era where digitization programs worldwide can point to the same unified standards defined in ISO 19264-1:2021 Photography — Archiving systems — Imaging systems quality analysis — Part 1: Reflective originals. This was a significant achievement in the evolution of cultural heritage digitization, which opens up the possibility of a truly universal body of digital knowledge available to anyone, anywhere, and at any time. This is the work of the ISO/TC42 JWG 26 team, comprised of experts selected by national standards bodies worldwide. Some of these experts are official representatives of the nations they represent; others are independent experts serving without compensation.

http://www.archives.gov/preservation/technical/guidelines.pdf https://creativecommons.org/publicdomain/zero/1.0/ iv

FADGI Code of Ethics for the Still Image Working Group The FADGI Still Image Working Group (http://www.digitizationguidelines.gov/still-image/) recognizes the code below, and this document follows this code.

FADGI Code of Ethics for Still Image Working Group -- January 24, 2020 The Federal Agency Digital Guideline Initiative (FADGI) Code of Ethics is a set of professional practices to guide imaging practitioners to create faithful reproductions of historic records held in the public trust. By conforming to the guidelines described by the FADGI program we believe values such as accuracy, integrity, and fidelity to the historic record are maintained. The cultural heritage community has a responsibility to produce digital images that look like the original records (textual, photograph, map, plan, etc.) and are a “reasonable reproduction” without enhancement. We encourage the use of digital technology to increase access to information without altering the fundamental nature of the historic record. We advocate for the dissemination of best practices to educate digital imaging practitioners. We recommend the inclusion of metadata or other documentation of the imaging process and any special processing performed on master digital objects.

The FADGI conformance program relies upon well-established practices, objective measurements, and quality assurance methods to create accurate imaging. The Program aims to reduce equipment variability, image artifacts, and other defects introduced by the scanning process and human subjectivity.

To be considered FADGI compliant, a digital imaging process must implement a digital imaging conformance evaluation program, and compliance should be demonstrated in project documentation or metadata.

Digitized material should document the appearance of the original at the time of capture, not what it may once have looked like if restored to its original condition. Traditional techniques such as contrast and brightness adjustments, dodging, burning, spotting, etc., may be allowed on digital images. Certain digital enhancements should be avoided such as: Colorization, color restoration, digital infill, excessive cropping of the image area, over sharpening, de-saturation, blurring, removal of blemishes in the original such as stains, dust, or scratches, and other transformative image processing. The alteration of cultural heritage images such as the addition, removal, or changing of image details, including the addition, changing, or removing of individuals, subject matter, scenery, or the unrealistic changing of color or light are not acceptable. A successful imaging product should reflect the completeness, image quality (tonality and color), and reflect the intellectual context of the original material.

Other practices such as multi-spectral imaging, contrast adjustments, or other implementation of technical tools that increase access to information of faded text, burnt materials, mold damaged, image stitching or other physical limitations are acceptable, but should have metadata or other documentation that describes the image processing.

It is recommended that a well-defined set of metadata be included either in project documentation or file metadata that clearly describe any imaging procedures that fall out of the normal imaging processing that alters the appearance of the original such as:

• Colorization

• Composite image

• Stitching

• Sharpening

• False Color

• Cropping

• Enhancement

• Saturation

• Desaturation (turning color to grayscale)

We recognize that the set of practices described in the FADGI guidelines only cover a small range of professional responsibilities within the cultural heritage field. These practices are intended to compliment and inform other related code of ethics, and do not supersede applicable statutory, legal, or other code of practices.

http://www.digitizationguidelines.gov/still-image/

Table of Contents

DOCUMENT INFORMATION ....................................................................................................................... II

Source Documents ..................................................................................................................................... ii

ACKNOWLEDGEMENTS ............................................................................................................................ III

FADGI CODE OF ETHICS FOR THE STILL IMAGE WORKING GROUP ................................................ IV

TABLE OF CONTENTS

CHAPTER 1: INTRODUCTION

1.1 Scope

1.2 Revisions

1.3 FADGI Glossary

CHAPTER 2: THE FADGI CONFORMANCE PROGRAM

2.1 The FADGI Star System

2.2 Components of FADGI Conformance

2.3 Digital Image Conformance Evaluation Process Monitoring

2.4 Evaluation Parameters

2.4.1 Sampling Frequency

2.4.2 Tone Response (OECF)

2.4.3 Gain Modulation

2.4.4 White Balance Error

2.4.5 Lightness Uniformity

2.4.6 Color (Encoding) Accuracy

2.4.7 Color Channel Misregistration

2.4.8 SFR* (Spatial Frequency Response)

2.4.9 Reproduction Scale Accuracy

2.4.10 Sharpening

2.4.11 Noise

2.4.12 Skew & Rotation

2.4.13 Geometric Distortion (Partially Implemented)

2.4.14 Field Artifacts (Future Implementation)

2.4.15 Highlight/Shadow, Tolerance

2.4.16 Dynamic Range

2.4.17 Color Management

2.5 FILE FORMATS

2.5.1 Master File Format

2.5.2 Access File Formats

2.5.3 File Compression

2.5.4 File Format Comparison

2.6 Physical Environment

2.6.1 Room

2.6.2 Monitor, Light Boxes, and Viewing Booths

2.6.3 Cleanliness of Work Area

2.6.4 Vibration

2.6.5 Flare

2.6.6 Lighting

2.6.7 Accessories

2.7 Limitations of the Guidelines

2.8 Resources

2.9 Summary

CHAPTER 3: EVALUATION CRITERIA VALUES FOR SPECIFIC MATERIAL TYPES

3.1 Bound Volumes: Rare and Special Materials

3.2 Bound Volumes: General Collections

3.3 Documents (Unbound): Manuscripts and Other Rare and Special Materials

3.4 Documents (Unbound): General Collections

3.5 Documents (Unbound): Modern Textual Records

3.6 Oversize Items: Maps, Posters, and Other Materials

3.7 Newspapers

3.8 Prints and Photographs

3.9 Photographic Transparencies: 35mm up to 4”x5”

3.10 Photographic Transparencies 4” x 5” and Larger

3.11 Photographic Negatives: 35mm up to 4”x5”

3.12 Photographic Negatives 4” x 5” and Larger

3.13 Paintings and Other Two-Dimensional Art (Other Than Prints)

3.14 X-Ray Film (Radiographs)

3.15 Printed Matter, Manuscripts, and Other Documents on Microfilm

CHAPTER 4: DIGITIZATION EQUIPMENT

4.1 Camera

4.2 Lens

4.3 Scanner

4.3.1 Planetary Scanner

4.3.2 Flatbed Scanner

4.3.3 Film Scanner

4.3.4 Drum Scanner

4.4 Selection of Digitization Equipment

CHAPTER 5: IMAGING WORKFLOW

5.1 Overview

5.1.1 Define the Requirement

5.1.2 Assess Organizational Capabilities

5.1.3 Insource vs. Outsource

5.1.4 Project Management

5.1.5 Workflow Plan

5.2 Large Scale Project Workflow

5.3 Small Scale Project Workflow

5.3.1 Sample Image Processing Workflow

CHAPTER 6: ADJUSTING IMAGE FILES

6.1 Overview

6.2 Color Correction and Tonal Adjustments

6.3 Cropping

6.4 Compensating for Minor Deficiencies

6.5 Stitching

6.6 Scanning Text

6.7 Optical Character Recognition

CHAPTER 7: TECHNICAL OVERVIEW

7.1 Spatial Resolution

7.2 Signal Resolution

7.3 Color Mode

7.4 Quantifying Scanner/Digital Camera Performance

7.4.1 Test Frequency and Equipment Variability

7.4.2 Reference Targets

CHAPTER 8: METADATA

8.1 Application Profiles

8.2 Data or Information Models

8.3 Levels of Description

8.4 Common Metadata Types

8.4.1 Descriptive

8.4.2 Administrative

8.4.3 Rights

8.4.4. Technical

8.4.5 Structural

8.4.6 Behavior

8.4.7 Preservation

8.4.8 Tracking

8.4.9 Meta-Metadata

8.5 Assessment of Metadata Needs for Imaging Projects

8.5.1 Relationships

8.5.2 Permanent and Temporary Metadata

CHAPTER 9: IDENTIFIERS AND FILE NAMING

9.1 File Naming

9.2 Directory Structure

9.3 Versioning

9.4 Naming Derivative Files

CHAPTER 10: QUALITY MANAGEMENT

10.1 Inspection of Digital Image Files

10.2 Quality Control of Metadata

10.3 Testing Results and Acceptance/Rejection

CHAPTER 11: STORAGE RECOMMENDATIONS

11.1 Digital Repositories and the Long-Term Management of Files and Metadata

APPENDIX A: GLOSSARY

Chapter 1: Introduction

Guidelines are important tools for maintaining standards and consistency in many fields. Cultural heritage imaging is no different. Guidelines provide the framework by which standards can be successfully approached. With standards across the output of the whole field, it is easier to have a unified approach to imaging and makes it easier to discuss and resolve issues as a community. The purpose of this document is to present a common set of guidelines to cultural heritage and digitization professionals from a wide range of organizations and levels of experience.

The 2023 revision of the Technical Guidelines for Digitizing Cultural Heritage Materials builds on more than two decades of shared experience of the cultural heritage imaging community worldwide. The guidelines presented in this document are in full alignment with the work of the International Organization for Standardization ISO/TC42 JWG 26, responsible for the development of the ISO 19246 standard that defines the metrics for professional cultural heritage imaging worldwide.

These guidelines are intended to be used with appropriate imaging targets and software designed to provide analysis of the digital image, which is an essential element of a FADGI-conforming imaging program. There are a variety of imaging targets and analysis software available worldwide. Any of these may be incorporated into a FADGI-conforming process if they meet the ISO 19264 standard and can provide all of the analysis functions required by FADGI.

The U.S. Library of Congress is the host institution for FADGI, and all Guidance provided within this document has been validated by the Library’s digitization program. We welcome professional visitors to visit our facilities on Capitol Hill.

1.1 Scope

These guidelines provide information on digitizing a wide range of materials. The evolving role of digitization in the preservation and presentation of knowledge has expanded our guidance in embracing that challenge. Some of these sections are a work in progress. These are labelled provisional.

These guidelines are informative, not prescriptive. Although they provide some practical advice, they are not meant to be a step-by-step manual for achieving results. No set of quality metrics or tests could ever replace the skill and experience of digitization professionals in evaluating the “goodness” of a digital representation of a physical item. However, the reverse is also true. Without a carefully designed and managed quality assurance program, no claim of FADGI conformance can be made.

The guidelines presented in this document provide a technical foundation for a FADGI-conforming digitization program, but do not exclude the adoption of alternate approaches that achieve conformance with the ISO 19264 standard.

The intended audience for these guidelines is cultural heritage digital imaging professionals and those who will be planning, managing and approving digitization projects. For those working on digital image capture and quality control, a foundation in photography and digital imaging is essential. Cultural Heritage digitization is a specialization within the imaging field that requires specific skills and experience. The FADGI Still Image Working Group has compiled these specific recommendations and best practices as they are practiced at participating institutions. Implementation of these recommendations must be managed by personnel with appropriate experience or in consultation with institutions or experts experienced in the FADGI digitization process.

1.2 Revisions

These guidelines reflect the current best practices within FADGI participating institutions and the world community of cultural heritage imaging professionals. These will change as technology, expectations, and applications for digital content evolve. We welcome your comments and suggestions at fadgistillimage@loc.gov.

The online version of the Guidelines is the official document, with addendums as listed on the FADGI website at www.digitizationguidelines.gov.

mailto:fadgistillimage@loc.gov http://www.digitizationguidelines.gov/

The following are the major revisions to the 2023 version of these guidelines:

• Addition of the FADGI Code of Ethics for the Still Image Working Group

• Addition of Professional Staff to the FADGI Conformance Program parameters

• Updates to the descriptions of the digital image conformance evaluation parameters

• Additional language on color management as an image conformance measure

• Addition of the Gain Modulation image conformance measure (informative)

• Addition of the Modern Textual Records material category with specifications and information provided by the National Archives and Records Administration

• Values for evaluation criteria presented as L*a*b* values rather than RGB 8-bit count values

• Removal of one-star specifications for Rare and Special collections material categories

• Updated values in the materials categories tables and the designation of certain criteria as

“informative” rather than essential evaluation criteria

• Updates to the Metadata section to reflect current standards

• Updates to the Adjusting Image Files (post-processing) section to reflect current best practices

• Updates to example digitization project workflows

• Addition of the Glossary section (Appendix A)

• Removal of the Resources and Technical Implementations sections – this content can now be found on the FADGI website linked to this page:

https://digitizationguidelines.gov/guidelines/digitize-technical.html

1.3 FADGI Glossary

See Appendix A of this document for a glossary of terms found in these guidelines. Please refer to the full FADGI Glossary for more extensive definitions of digitization terminology at http://www.digitizationguidelines.gov/glossary.php. Note that the online glossary definitions are the most up-to-date.

Also see ISO 19262 for additional detail at https://www.iso.org/standard/64219.html.

Chapter 2: The FADGI Conformance Program

2.1 The FADGI Star System

The International Organization for Standardization ISO 19264 document defines image quality in a three step system that includes Levels A, B and C. Nominally, FADGI corresponds to these three levels as Four Star (Level A or best), Three Star (Level B or good) and Two Star (Level C or acceptable). Additionally, FADGI provides guidance for a One Star conformance for some imaging tasks. One-star imaging does not conform to the ISO 19264 standard, and is offered by FADGI for reference use only. See the descriptions of the four FADGI Star levels below.

The FADGI process places high importance on image quality consistency. Individual guidance for each subset of content type may place greater importance on one or more of the measured parameters, and may not measure some of the parameters at all. This is an important refinement to the ISO 19264 metrics which takes into account the type of materials being imaged, and the intended use of the digital images being created.

Earlier versions of the FADGI Guidelines used RGB 8-bit count values and other measurement methods that are no longer appropriate. As a result, you will find that throughout these guidelines, measurements are changed to L*a*b*. The earlier count value guidance is still present in some places for reference, but will not precisely match the L*a*b* values. Earlier values do not validate FADGI conformance for newly-produced images (following the publication of this document) and are no longer valid for 2023 version conformance.

A digital object’s star rating is determined by meeting the FADGI evaluation parameters through image analysis. A narrative description of the star system ratings is summarized below:

https://digitizationguidelines.gov/guidelines/digitize-technical.html http://www.digitizationguidelines.gov/glossary.php https://www.iso.org/standard/64219.html

One-Star imaging is appropriate for applications where the intent is to provide a reference to locate the original or when there is no ability to image to a higher star level. FADGI recognizes that there are many reasons why imaging to higher star levels is not possible, but conformance to FADGI guidance for other aspects of an imaging program can be highly beneficial. A one-star rating is an excellent starting point for any program that wishes to begin standardized imaging but may not have the necessary support or experience to achieve a higher rating.

Two-Star imaging is appropriate where there is no reasonable expectation of having the capability to achieve three-star conformance, or when three-star conformance is not needed for the intended use of the digital image. Two-star conformance is the minimally acceptable rating for most professional digitization activity.

Three-Star imaging is defined as producing a very good professional image that is appropriate for most uses.

Four-Star defines the best imaging practical today. Images created at the FADGI four-star level represent the current state of the art in image capture. Generally speaking, the four-star level is an appropriate goal for smaller scope projects as it tends to be more achievable.

Our mission is to define what is practical and achievable today, and provide you with the knowledge and tools appropriate for your program. Generally, in order to avoid future rescanning and given the high cost and effort involved in digitization, FADGI does not recommend digitization to less than a three-star level when possible, and cautions against imaging to four-star level for all but the most sophisticated imaging programs. The FADGI program balances the desire for quality with the limitations of individual digitization programs.

2.2 Components of FADGI Conformance

There are four primary components of a FADGI compliant digitization program:

Technical Guidelines and Imaging Parameters This consists of the guidance included in this document as well as other guidance that may be appropriate from a variety of sources, such as the International Organization for Standardization (ISO), equipment manufacturers, professional associations, and imaging consultants. No one source of information is comprehensive or exclusive.

Best Practices These are documented workflows and processes that include guidance on addressing common challenges or situations, which may be unique to your program or follow established workflows developed by others. These should be based on a continuous improvement process methodology.

Digital Imaging Conformance Evaluation All FADGI-conforming digitization programs must incorporate repeatable and reliable methods of validating the consistency and accuracy of imaging “goodness” as defined in this document for the various types of materials covered in this document. Once conformance is established, it’s important to not make changes to digitization workflows or production environments. Testing should be conducted any time that any conditions changed.

For imaging tasks outside of the scope of this document, appropriate methods of measuring and monitoring production must be developed and incorporated into the production workflow. FADGI would appreciate your input for additional imaging processes to be incorporated in future editions of these guidelines, or as addendums to this document as appropriate.

Professional Staff FADGI compliant digitization is a complex, highly technical process that requires skillsets spanning a number of separate but related fields. Beyond the need for understanding photographic and digital imaging processes, staff need training in proper care and handling of materials and need an understanding of the total lifecycle of the digitization program. No claim of FADGI conformance can be made for programs that do not have appropriately trained staff and do not provide ongoing staff professional development.

These elements, when implemented together, form a FADGI-compliant digitization environment.

2.3 Digital Image Conformance Evaluation Process Monitoring

A digital image conformance evaluation program provides the measurement and monitoring component of a FADGI-compliant digitization program. Such a program consists of two components:

• Image Targets, both reflective and transmissive

• Analysis Software Imaging targets have been designed to comply with various ISO specifications, and the parameters as defined in the FADGI program have been validated through years of use at participating Federal agencies.

It is beyond the scope of our mission to validate or recommend targets and software, however, they may be incorporated into professional digitization programs which meet the ISO 19264 standard and can therefore claim conformance to the ISO standard. Please refer to the FADGI website Resources page for examples of targets and software that are appropriate for FADGI-conforming imaging programs.

2.4 Evaluation Parameters

The FADGI guidelines establish quality and performance goals for the four levels of the star ranking system. Digital image conformance testing tools, when used with appropriate testing targets, provide the user with precise and repeatable analysis of the imaging variables that comprise FADGI star ratings. As noted above, measurements are now changed to L*a*b*. Count values from earlier versions of these guidelines have been removed. Please see the Resources page for these guidelines for more information on these changes (https://www.digitizationguidelines.gov/guidelines/digitize-technical-resources.html).

Certain evaluation parameters are strictly informative and do not validate FADGI conformance for newly-produced images. They appear as white rows with light gray text in the tables with measurement parameters in Chapter 3.

The parameters used to determine FADGI conformance are described below.

2.4.1 Sampling Frequency

This parameter measures the potential spatial resolution of a digital image, and is computed as the physical pixel count in pixels per inch (ppi) or pixels per millimeter. Ultimately, this parameter informs us about the size of the original and also provides part of the data needed to determine the potential level of detail recorded in the file. ISO 12233:2014 defines the resolution measurements.

The Resolution (Sampling Frequency) criteria in the tables for each material category (Chapter 3, beginning on page 27) include a tolerance for the Reproduction Scale Accuracy percentage. Please see the description of that criteria below for more information on the relationship between the two parameters.

The tables in the following chapter include the minimum sampling frequency in ppi for each content category to meet each star level of FADGI compliance. It’s important to analyze the source material and select a resolution that’s appropriate based on the physical attributes of the item, as well as the relationship between the item’s size and the level of detail desired.

https://www.digitizationguidelines.gov/guidelines/digitize-technical-resources.html

2.4.2 Tone Response (OECF)

Opto-Electronic Conversion Function (OECF) is a description of how the digital imaging system converts light into digital values. ISO 14524:2009 defines the OECF measurement. Most imaging systems for consumer and professional applications “optimize” the tone response to create a visually appealing representation of the scene. This is commonly known as an “S” curve, where the midtone of the image is straight and both highlight and shadow portions of the scene are rendered for a more pleasing image.

These optimizations are not standardized between manufacturers and pose challenges to restore to the linear tone response essential for cultural heritage digitization. The measurement for this metric is made in colorimetric L* units (noted as ΔL2000* where Δ represents the delta).

Note: The Tone Response (OECF) values for transmissive material types in the following chapter have been determined to best capture the tones in the original transmissive item (film, negatives, etc.). Any other adjustments to the image will be done during post-processing. These guidelines focus on accurate capture of the transmissive original, while other adjustments to create a more “pleasing” derivative may be done after capture. The values for transmissive collection types are considered provisional. This is indicated in the evaluation parameters tables in Sections 3.9, 3.10, 3.11, 3.12, and 3.14.

2.4.3 Gain Modulation

The gain modulation is a measure of the localized slope of the OECF (tone response) when plotted as Lout* (y-axis) versus Lin*(x-axis). Its primary benefit is to objectively measure and detect clipping or soft clipping distortion behaviors by way of the OECF or tone scale curve. Ideally the slope of the OECF curve should be sufficiently high to make Lin* values easily recoverable from the Lout* values without introducing quantization or posterization artifacts. If the slope of the above L* curve is too low, deriving original Lin* values from captured Lout* values becomes ambiguous, especially in the presence of image noise. Most clipping behaviors occur at the extreme values of the Lout* vs. Lin* curve.

It should be noted that the upper and lower L* tolerances for 3- and 4-star levels often catch such clipping behaviors by themselves. This is typically not the case for 1- and 2-star levels though. This is where the gain modulation metric offers an additional advantage.

Note: In this revision, this metric is informative only, and not required for evaluating whether an image is FADGI-compliant. As such the rows for this metric are shaded white to differentiate them from the gray rows with the required metrics in the tables in the next chapter.

2.4.4 White Balance Error

This is a measurement of the color neutrality of the digital file. The definition of neutral is not universal.

RGB workflows that use digital count values encode neutral as defined by the International Color Consortium (ICC) color space chosen. L*a*b* workflows define neutral as 0 on the a* axis and b* axis, with the lightness recorded from 0-100 on the L* axis. Digitization workflows are increasingly adopting the L*a*b* color mode. It is worthwhile noting that the gray patches used for measuring white balance will never truly have a* and b* values of zero. Typically, these values are sufficiently small to ignore in day-to-day workflows. Nevertheless, these small values can be accommodated for in reporting colorimetric white balance by removing the slight bias.

There are several other color models that are used by various applications. Key to successful FADGI conformance is the establishment of the correct white balance before beginning the ICC calibration process. The metric for colorimetric evaluation of white balance is similar to the color accuracy ΔE2000 metric, with the exception that the L* component is eliminated. The resulting metric is defined as ΔE(a*b*).

Note: The values for transmissive collection types are considered provisional. This is indicated in the evaluation parameters tables in Sections 3.9, 3.10, 3.11, 3.12, and 3.14.

2.4.5 Lightness Uniformity

It is not possible to create a perfectly even field of illumination without the use of software that takes into account the inherent fall-off of all lenses and the limitations of lighting systems. Ideally, there should be a perfectly even recording of a neutral reference from center to edge and between points within the image.

https://en.wikipedia.org/wiki/Clipping_(audio) https://www.hackaudio.com/digital-signal-processing/distortion-effects/soft-clipping/ https://www.hackaudio.com/digital-signal-processing/distortion-effects/soft-clipping/

ISO 17957:2015 defines the shading measurements. Specific values are defined in each section of the guidelines. Flat fielding software using captured data from extended and uniform matte surface targets provides a simple and effective method of reducing non-uniformity, but should not be a substitute for good professional imaging technique. Aggressive corrections can in themselves introduce digital noise artifacts into the finished image file. Target surfaces with glossy or semi-gloss surfaces should be avoided.

Note: The values for transmissive collection types are considered provisional. This is indicated in the evaluation parameters tables in Sections 3.9, 3.10, 3.11, 3.12, and 3.14.

2.4.6 Color (Encoding) Accuracy

True color accuracy can only be evaluated once a digital image has been rendered as a physically measurable color to print or display. It is important to understand that a digital image, in the context of FADGI, is an encoding of color as delivered by any number of digital image file formats. We treat color accuracy here as if the color code values are idealized virtual colors as interpreted by assigned color profiles or color working spaces. These color profiles and color spaces can be considered color dictionaries for translating image code values into idealized rendered colors. For simplicity in communication, we continue to describe this metric as ‘color accuracy’ with the understanding that doing so requires the use a working color space dictionary for interpretation of RGB values to standardized working color space.

There is no perfect imaging system or perfect method of color evaluation. Color accuracy is measured by computing the color difference (ΔE2000) between the decoded imaging results of an assigned color working space to the pre-measured color values of a color target. By imaging an appropriate target and evaluating it through software, variances from known values can be determined, which is a good indicator of how accurately the system is recording color. Software programs generally measure the individual patch values and also provide the average deviation of all color patches measured (the mean). Refer to ISO 13658:2000 for additional documentation on color accuracy measurement.

As a general rule, targets with a small number of color and density patches can provide a good analysis of system consistency, but cannot provide information suitable for creating ICC color profiles. Targets with 100 or more measurement patches will provide a better input to ICC profile creation software when making ICC camera (input) or master file (output) profiles.

Profiling for printer calibration is done using targets generated by the ICC profiling software in use and printed on the specific device being calibrated. This involves the measurement of hundreds or thousands of printed reference patches, which are then compared to the values in the digital file being printed.

Printers generally have a far less linear response requiring more data points to properly calibrate. Digital image conformance evaluation targets are not appropriate for the calibration of printers.

A note of caution: there are many manufacturers of devices which measure reflective and transmissive color. However, each manufacturer has their own standard for calibration of their instruments, and measurements can vary between properly calibrated devices from different manufacturers. For the most consistent results imaging devices should be re-certified on a rotating schedule, with the before and after recertification values recorded for comparison. Take care to use the same target for both before and after recertification, ideally a target designed to be a stable reference. Target and device variability are a primary cause of inconsistent imaging stability.

2.4.7 Color Channel Misregistration

All lenses focus red, green, and blue light slightly imperfectly. This parameter measures the spatial dislocation of red, green, and blue light in terms of pixels of misregistration. This parameter is used in the evaluation of lens performance. Poor registration of the three color channels can indicate equipment issues, poor conditions, etc. Alternate terms used for this metric are color fringing, chromatic aberration, and lateral color error, which all fall into the general category of lens aberrations.

2.4.8 SFR* (Spatial Frequency Response)

*Note: The SFR is a surrogate metric to Modulation Transfer Function (MTF). There are some very minor differences between SFR and MTF. In the purest sense SFR may not give the same results due to assumptions about linearity and the response at the zero frequency. For simplicity we continue to use SFR as our metric in this document. For field use SFR is much easier to calculate.

It is related to the sharpness, focus, and ability to resolve fine details. The SFR is a functional (i.e. multi-valued) relationship, which defines an imaging system’s ability to maintain contrast of increasing smaller details. Figure 1 (below) illustrates this. There are several single valued metrics to describe the SFR function. These include measurements at specific points on the curve, such as the 50% and 10% response points as well as the response at the highest frequency supported by the sampling frequency.

ISO 12233:2000, ISO 16067-1:2003, and ISO 16067-2:2004 define SFR measurement.

Since the Spatial Frequency Response (SFR) is a functional relationship with many data points, it is difficult to specify an absolute curve shape unequivocally. Therefore, single numbered metrics have been developed to summarize the important characteristics of the SFR. These include, SFR50, SFR10 (Sampling Efficiency), SFR Response at Nyquist Frequency, Max SFR, and SFR Abnormal Behavior (under development). A description of each of these metrics follows:

• SFR50: This is a measurement which determines the frequency when the spatial frequency response (SFR) has a value of 50% modulation. That is, only 50% of the original content contrast has been maintained for that level of detail. In order for this frequency to be compared across images of any arbitrary sampling frequency, e.g. 300, 400, 600 ppi, the value is normalized to the half sampling (Nyquist) frequency and expressed as a percentage. For example, if an image captured at 400 ppi has an SFR with 50% modulation at the 120 ppi level of detail, then the SFR50 metric is expressed as 60% [(120 / 200) *100]. Note: The half sampling (Nyquist) frequency is half of 400 ppi or 200 cycles/inch. This is because it takes two pixels to define a cycle. The formula relating ppi to cycles per inch is shown below.

• SFR10 (Sampling Efficiency): This is a measurement which determines the frequency when the spatial frequency response (SFR) has a value of 10% modulation. Again, the measured frequency is normalized to the half sampling (Nyquist) frequency and expressed as a percentage. For example, if the SFR curve has 10% modulation at 180 ppi for an image captured at 400 ppi, the sampling efficiency would be expressed as 90% [(180 / 200) *100].

• SFR Response at Nyquist Frequency: This is a measurement of the spatial frequency response at exactly the half sampling (Nyquist) frequency. Note: In this revision, this metric is informative only, and not critical for evaluating whether an image is FADGI-compliant. As such the rows for this metric are shaded differently than the critical metrics in the tables in the following sections.

Figure 1 : Graphical Illustration of SFR using a resolution test pattern imaged by a lens showing loss of contrast with increasing spatial frequency

• SFR Abnormal Behavior: This is a metric to be developed, which will try to identify and characterize non-typical SFR shapes, most likely caused by some post-processing. Since it’s still under development, it’s not included in the evaluation parameter tables in the following chapter.

• Sharpening (Max SFR): This is a measurement of the maximum value of the SFR function. If the modulation exceeds 1.0 (100%), it can be an indication of digital enhancement filtering. Figure 2 at right shows in more detail how contrast is reduced at higher spatial frequencies.

2.4.9 Reproduction Scale Accuracy

This parameter measures the relationship between the size of the original object and the size of that object as calculated from the digital image metadata. This parameter is measured in relation to the pixels per inch (ppi) or pixels per millimeter (ppmm) of the original digital capture. For example, capturing an image of a ruler at 400 ppi will digitally render the ruler at the correct size when displayed or printed at 400 ppi. It is critically important in cultural heritage imaging to maintain the relationship to the original size of the object. Too often, accurately populating the metadata with the correct ppi at capture is treated casually.

Calculating the reproduction scale accuracy requires the resolution value defined in the header of the image file. This value becomes the target/reference resolution that the image content is measured against. If this information is missing from the header of the image file, the image is not FADGI compliant.

The original size of microfilmed documents can only be determined if the filming “reduction ratio” is known, or there is a physical ruler in the image area, which is not normal practice. The reduction ratio is referred to as 8x or 10x (or other) reduction, indicating that the magnification of the image on the microfilm is 1/8th or 1/10th the size of the original document. This may or may not be known when digitizing microfilm. Unless noted in metadata, the scale of the original will be lost when microfilm is digitized. Microfilm is digitized at a fixed ppi resolution, regardless of the original “reduction ratio.”

Photographic film cannot be related to a reproduction scale, unless there is a physical measurement in the image to scale to. Photographic film is digitized to appropriate resolutions relative to the size of the film and the capabilities of the digitization system.

2.4.10 Sharpening

Almost all digital imaging systems apply sharpening, often at a point where the user has no control over the process. Sharpening artificially enhances details contrast to create the illusion of greater definition.

Digital image conformance evaluation testing quantifies the level of sharpening present. There are three major sharpening processes in a typical imaging pipeline: capture sharpening (through camera setting

Figure 2: Light intensity cross sections of resolution wedge at different ppi line spacing adjustment), image sharpening in post processing, and output sharpening for print or display purposes.

Sharpening is usually implemented through image edge enhancement, such as filtering techniques using unsharp masks and inverse image diffusion. There is nothing inherently wrong with sharpening files at any of these three steps, unless doing so decreases the image quality “goodness”. Modern digital cameras and scanners often apply a significant amount of image processing to captures by default, much of which cannot be turned off. However, sharpening is irreversible, and best practice for archival master files is to limit the amount of sharpening applied to master files.

The examples in Figure 3 below show the visual and SFR differences between a sharpened but low resolution image on the left, compared to an unsharpened but high resolution image on the right. Once the SFR of the left image is below the 10% level (i.e. 90% contrast loss) there is no way to reliably recover the higher resolution information seen in the right image. Notice the inset and enlarged portions of the clock feature. The minute marks are much clearer in the unsharpened image.

As noted above, the lower frequencies of the SFR of the left image in Figure 3 have been selectively enhanced in contrast via image processing. Figure 4 (right) shows a graphical representation of the relationship between resolution and sharpness.

2.4.11 Noise

Generally, noise can be defined as any small scale or pixel-to-pixel variability in L* or RGB count values.

A good example is the kind of granular deviation or grain (i.e. noise) seen in photographic film. In other words, any unexpected count value or L* deviations at the pixel-to-pixel level. This requires a test target feature with very low texture or grain itself. Very simply, noise is calculated as the standard deviation of L* values within the region of interest of a given gray patch.

Noise in digital images can take on a whole new meaning because of image processing artifacts. For instance, standard JPEG processing introduces 8x8 pixel blocks that are clearly not part of the original content. In many cases camera/scanner user interfaces often allow for some form of noise cancelling or de-noising. As with sharpening selections, users should use these in moderation. Aggressive de-noising selection can completely remove any naturally occurring texture in a scanned object leading to unnatural appearing images. For instance, the texture or tooth in parchment or textured papers may be completely removed. This is why we have inserted cautionary alerts in the FADGI noise metric sections for any noise values less than an indicated value. Noise values that are too low are a possible sign of aggressive de-noising selections

Figure 3: Example difference between sharpening (left) and resolution (right)

Figure 4

Noise levels will also increase with aggressive sharpening selections, high ISO speed settings, and short exposure times. Noise cannot be measured using a target with any kind of surface texture. The target should have a smooth or glossy surface.

2.4.12 Skew & Rotation

Skew and rotation are similar geometric distortions but have distinctly different characteristics. Examples of each are shown below in Figure 5. The middle image shows the starting content. The page content is rectilinear and the sides are perfectly horizontal and vertical with 90° angles between all adjoining sides.

The left image is a skewed version. None of the intersecting angles are at 90°. Even if one of the angles was not at 90°, it would still be considered skewed. Skew behaviors in the captured image can often be traced to lack of parallelism between the focal plane of the camera and the platen.

Rotation keeps all the parallel features of the opposite sides and maintains 90° angles at all corners and is introduced at capture by improper content alignment at capture.

Figure 5 These parameters measure how straight the image is in the file. Skew and rotation are the angle of deviation in a digital image from the paper edge, text lines or other visual reference elements of the original item. Master image files should be straight in relation to the contents of the image. The guidelines now allow for rotation correction to be applied to images.

2.4.13 Geometric Distortion (Partially Implemented)

Critically important to faithful reproduction of an original is the management of geometric distortion in image capture. Typical camera lenses are poorly corrected for this, and create images which exhibit significant distortion even under ideal conditions. If the imaging system is not correctly aligned, other distortions are introduced to the image as well. Cultural heritage imaging requires high quality optics designed for minimal geometric distortion. Recently digital correction processes have been introduced to “correct” for the distortions created by lenses from many manufacturers. While these software approaches are interesting and often effective, they interpolate pixels which causes loss of image integrity. High quality lenses designed for copy work will generally have very well-controlled geometric distortion, and should not be corrected further through software for master files. ISO 17850:2015 defines the geometric distortion measurements for digital cameras.

2.4.14 Field Artifacts (Future Implementation)

Ideally, digitization should only capture what is in the original object. However, dust, dirt, and scratches almost inevitably find their way into digital files. This parameter quantifies the physical non-image artifacts in a digitization system. It should be noted that this parameter refers to artifacts ON the object, not other artifacts like sensor dust. Current best practice involves post-capture physical inspection for non-image artifacts, and noting their frequency to inform the appropriate level of post-process inspection. Editing out artifacts should only be done to derivative files.

2.4.15 Highlight/Shadow, Tolerance

Unlike for reflective materials, where the colorimetry can be scene-referred, for transparencies the true colorimetry of the original scene is lost. Instead we only have a proxy to the original scene, i.e. the original scene imaged onto the film. This is an emerging area and the final values are provisional/informational.

2.4.16 Dynamic Range

Dynamic range is the difference of the minimum and maximum density values an imaging system can capture (D max minus D min). The dynamic range is calculated for both capture systems and source material. This is an emerging area and the final values are informative only.

2.4.17 Color Management

As of the date of this publication, ICC color management is the most widely used process of attempting to correct for the many variables inherent in the digital imaging process. Each component of the image capture cycle influences the rendering of the digital image, and no two image capture systems are identical. There are many, many approaches to resolve these differences, and just as in the capture process itself, most do not achieve the same result given the same input.

As a result, we still cannot achieve consistently similar imaging using a mixed bag of imaging systems and color calibration methods and software.

There is no fix to this issue if device-level color management solutions are used. Unless all of the imaging conditions, including cameras/scanners/software/lighting/lenses etc.