image analysis, sediments and paleoenvironments developments in paleoenvironmental research

143 Introduction Wavelet analysis Image processing Methodology Testing of the method Example: Marine Laminated sediments from the west coast of Vancouver Island, NE Pacific Advantages of

Image Analysis, Sediments and Paleoenvironments

Developments in Paleoenvironmental Research

VOLUME 7

Image Analysis, Sediments and

Paleoenvironments

Edited by

Pierre Francus

Springer

eBook ISBN: 1-4020-2122-4

Print ISBN: 1-4020-2061-9

©2005 Springer Science + Business Media, Inc

Print ©2004 Springer

All rights reserved

No part of this eBook may be reproduced or transmitted in any form or by any means, electronic,mechanical, recording, or otherwise, without written consent from the Publisher

Created in the United States of America

Visit Springer's eBookstore at: http://ebooks.springerlink.com

and the Springer Global Website Online at: http://www.springeronline.com

Dordrecht

I dedicate this book to my wife, Sophie Magos

This page intentionally left blank

Table of Contents

Editors and Board of Advisors of Developments in Paleoenvironmental

List of Contributors

1 An introduction to image analysis, sediments and paleoenvironments

Pierre Francus, Raymond S Bradley and Jürgen W Thurow 1

Part I: Getting started with Imaging Techniques

(or methodological introduction)

2 Image acquisition

Scott F Lamoureux and Jörg Bollmann 11

Introduction

Image acquisition and paleoenvironmental research

Sample preparation for image acquisition

Acquisition methods

Summary

Acknowledgments

References

3 Image calibration, filtering and processing

Alexandra J Nederbragt, Pierre Francus, Jörg Bollmann

and Michael J Soreghan 35

The Editor xii

Aims & Scope of Developments in Paleoenvironmental Research Book Series ix ii

xv

Research Book Series xiv

4 Image measurements

Eric Pirard 59

Introduction

Digital imaging and sampling theory

Dealing with the available information

Digital image analysis strategies

Intensity and color analysis

5 Testing for sources of errors in quantitative image analysis

Pierre Francus and Eric Pirard 87

Part II: Application of Imaging Techniques

on Macro- and Microscopic Samples

6 Digital sediment colour analysis as a method to obtain high

resolution climate proxy records

Alexandra J Nederbragt and Jürgen W Thurow 105

Introduction

Image data collection

Extracting colour data

7 Toward a non-linear grayscale calibration method for legacy

photographic collections

Joseph D Ortiz and Suzanne O’Connell 125

Introduction

What is grayscale analysis?

Evaluating the nonlinear correction method

Summary

Acknowledgments

Metadata

References

8 From depth scale to time scale: transforming sediment image color

data into a high-resolution time series

Andreas Prokoph and R Timothy Patterson 143

Introduction

Wavelet analysis

Image processing

Methodology

Testing of the method

Example: Marine Laminated sediments from the west

coast of Vancouver Island, NE Pacific

Advantages of using image analysis

Drawbacks to image analysis

Example: case study of five diamicton units

from North Atlantic continental margins

10 Application of X-ray radiography and densitometry in varve analysis

11 Processing backscattered electron digital images of thin sections

Michael J Soreghan and Pierre Francus 203

Introduction

Image acquisition

Image processing

Image measurement

Case study: grain size analysis of upper Paleozoic loessites

Discussion and recommendations for BSE image analysis

Part III: Advanced Techniques

12 Automated particle analysis: calcareous microfossils

Jörg Bollmann, Patrick S Quinn, Miguel Vela, Bernhard Brabec,

Siegfried Brechner, Mara Y Cortés, Heinz Hilbrecht, Daniela N Schmidt,

13 Software aspects of automated recognition of particles:

the example of pollen

Ian France, A W G Duller and G A T Duller 253

AIMS AND SCOPE OF

DEVELOPMENTS IN PALEOENVIRONMENTAL RESEARCH

SERIES

Paleoenvironmental research continues to enjoy tremendous interest and progress in the scientific community The overall aims and scope of the

Developments in Paleoenvironmental Research book series is to capture this

excitement and document these developments Volumes related to any aspect of paleoenvironmental research, encompassing any time period, are within the scope of the series For example, relevant topics include studies focused on terrestrial, peatland, lacustrine, riverine, estuarine, and marine systems, ice cores, cave deposits, palynology, isotopes, geochemistry, sedimentology, paleontology, etc Methodological and taxonomic volumes relevant to paleoenvironmental research are also encouraged The series will include edited volumes on a particular subject, geographic region, or time period, conference and workshop proceedings, as well as monographs Prospective authors and/or editors should consult the series editors for more details The series editors also welcome any comments or suggestions for future volumes

xiii

EDITORS AND BOARD OF ADVISORS OF

DEVELOPMENTS IN PALEOENVIRONMENTAL RESEARCH BOOK SERIES

LIST OF CONTRIBUTORS

JÖRG BOLLMANN

Department of Earth Sciences

ETH and University Zurich

Sonneggstrasse 5, 8092 Zurich

Switzerland

e-mail: bolle@erdw.ethz.ch

BERNHARD BRABEC

Department of Earth Sciences

ETH and University Zurich

Department of Earth Sciences

ETH and University Zurich

Sonneggstrasse 5, 8092 Zurich

Switzerland

MARA Y CORTÉS

Department of Earth Sciences

ETH and University Zurich

108 Walcot Street, Bath

BA1 5BG, United Kingdom

e-mail: andyduller@hotmail.com

G.A.T DULLER (Geoff.Duller@aber.ac.uk)

Institute of Geography and Earth Sciences,

Department of Earth Sciences

ETH and University Zurich

Department of Geological Sciences,

University College London,

ANTTI E.K OJALA

Geological Survey of Finland

Department of Earth Sciences

ETH and University Zurich

Kent State University

Lincoln and Summit Streets

Kent, OH 44224, USA

e-mail: jortiz@kent.edu

R TIMOTHY PATTERSON

Department of Earth Sciences and Ottawa-Carleton Geoscience Centre,

Herzberg Building, Carleton University

Ottawa, Ontario

K1S 5B6, Canada

e-mail: tpatters@ccs.carleton.ca

xvii

HANS R THIERSTEIN

Department of Earth Sciences ETH and University Zurich

Sonneggstrasse 5, 8092 Zurich Switzerland

e-mail: thierstein@erdw.ethz.ch JÜRGEN W THUROW

Department of Geological Sciences University College London

Gower Street, London WC1E 6BT, UKe-mail: j.thurow@ucl.ac.uk

MIGUEL VELA

Department of Earth Sciences ETH and University Zurich

Sonneggstrasse 5, 8092 Zurich Switzerland

e-mail: d.schmidt@gl.rhul.ac.uk MICHAEL J SOREGHAN

School of Geology and Geophysics, University of Oklahoma,

e-mail: schiebel@erdw.ethz.ch xviii

100 E Boyd St

e-mail: msoreg@ou.edu

é

1 AN INTRODUCTION TO IMAGE ANALYSIS, SEDIMENTS

AND PALEOENVIRONMENTS

PIERRE FRANCUS (pierre_francus@inrs-ete.uquebec.ca)

Climate System Research Center

RAYMOND S BRADLEY (rbradley@geo.umass.edu)

Climate System Research Center

Department of Geosciences

University of Massachusetts

Amherst, MA 01003-9297

USA

JÜRGEN THUROW (j.thurow@ucl.ac.uk)

Department of Geological Sciences

University College London

Gower Street, London WC1E 6BT

UK

Keywords: Visual information, Quantification, Geosciences, Image acquisition, Image processing, Image

mea-surement, Quality control, Neural networks, Recommendations

Image analysis is concerned with the extraction of quantitative information from imagescaptured in digital form (Fortey 1995) Visual information has always played an impor-tant role in the Geosciences — indeed, many disciplines rely heavily on the content ofimages, whether they are sketches drawn in the field, or descriptions of microscopic slides(Jongmans et al 2001) Visual charts are often used in sedimentology in order to providesome semi-quantification, such as for instance, Krumbein’s grain roundness classes (Krum-bein 1941), classification of ichnofabric (Droser and Bottjer 1986), or simply the chart ofphase percentages sitting nearby every binocular microscope However, with the noticeable

2 FRANCUS, BRADLEY AND THUROW

exception of remote sensing, compared to other disciplines image analysis has been slow

to develop in the Geosciences, despite its potential usefulness One problem with imageanalysis studies of geologic material is that objects are generally less homogenous thanbiologic or medical samples, and observation conditions are more variable

Digital imaging systems were the exception in the 80’s, because the computers needed toprocess sizeable images were cutting edge and expensive systems, mostly entirely tailoredfor that unique purpose The decreasing price of personal computers, with their simultane-ous and dramatic increase in performance, made digital image processing more accessible

to researchers in the 90’s Soil scientists, especially micromorphologists, have been veryactive in the development of new image analysis tools (e.g., Terribile and Fitzpatrick (1992),VandenBygaart and Protz (1999), Adderley et al (2002)) The growing interest for imageanalysis in Earth Sciences is revealed by the increasing number of initiatives to bringimage analysis into the spotlight Without being exhaustive, one can mention a number

of meetings on the subject (e.g., Geological Society of London, London, UK, September

1993, and Geovision held in Liège, Belgium, in May 1999), an increasing number of papers

in journals such as Computers & Geosciences, and books (e.g., De Paor (1996)) In thesecond volume of the Developments in Paleoenvironmental Research (DPER) series, achapter by Saarinen and Pettersen (2001) was already devoted to image analysis applied topaleolimnology

Paleoenvironmental studies of sediments can greatly benefit from image analysis niques Because it is a low cost and high-resolution analysis method, image analysis allowssediment cores to be studied at the very high resolution that is necessary to resolve highfrequency climate cycles For instance, image analysis of varved sediments can contribute

tech-to a better understanding of past climate variability, providing that chronologies are verifiedand quantitative relationships are established between the sedimentary record and climate

A wide range of data can be acquired using image analysis Visual data include counting

of laminations (to build-up time scale), measurement of lamination thickness, and lishment of sediment properties (chemistry, mineralogy, density) from its color Physicaldata are for instance the morphometry of microfossils such as diatom and coccoliths, grainsize, grain morphometry, sediment fabric Chemical and mineralogical data can be inferredfrom images of tools such as XRF-Scanning, IR-Scanning, and energy and wavelengthdispersive spectrometry Other tools used are X-radiography, core scanning, non-normalscanning, optical and electron microscopy

estab-An international group of scientists, mainly marine and lacustrine sedimentologists,gathered at the University of Massachusetts, Amherst, in November 2001 to review thissubject, and to make an update of the latest techniques available The workshop entitled

Image Analysis: technical advances in extracting quantitative data for paleoclimate construction from marine and lacustrine sequences was sponsored by the US-National

re-Science Foundation (NSF) and the International Marine Past Global Change Study AGES) program The participants of the workshop made recommendations (documented

(IM-in the appendix) promot(IM-ing the use of low cost image analysis techniques and facilitat(IM-ingintercomparisons of measurements in the paleoclimate community

This volume is the natural extension — not the proceedings — of the workshopbecause it addresses some of the concerns and fulfils some of the needs identified dur-ing the workshop Although image analysis techniques are simple, many colleagueshave been discouraged in using them because of the difficulty in gathering relevant

AN INTRODUCTION TO IMAGE ANALYSIS . 3information in order to set-up protocols and methodologies to solve a particular issue.Often, specialized papers are only comprehensible by computer scientists, mathematicians

or engineers Relevant information is scattered in the methods sections of many differentresearch papers, and is not detailed enough to be helpful for beginners Also, monographs

on image analysis techniques (e.g., Russ (1999)) are oriented towards medicine, biology ormaterial science Finally, specialized lectures remain very expensive The DPER volume

7 intends to fill this gap, providing comprehensive but simple information on imagingtechniques for paleoenvironmental reconstruction in a single volume By providing suchinformation, the user will understand every step involved in the imaging process, fromthe acquisition to measurements, in order to be able to evaluate the validity of scientificresults obtained This is necessary in order to allow image analysis techniques to mature

as widely accepted methodologies for paleoenvironmental reconstructions In brief, thisvolume intends to:

- provide a compendium of image analysis techniques available for paleoenvironmentalreconstruction retrieved mainly from lacustrine and marine sediment cores;

- cover image analysis techniques performed at the core-scale level (macroscopic,sedimentary structure, color), and at the microscopic-scale (thin-section, and X-ray slabs);

- provide comprehensive descriptions of protocols, guidelines, and recommendationsfor pertinent use of low cost image analysis techniques;

- review and illustrate the wide range of quantitative information that can be obtainedusing image analysis techniques by showing case studies;

- show improvements that high-resolution studies using image analysis techniques canbring about in paleoenvironmental reconstructions and in our understanding of environ-mental changes

In order to achieve these goals, the DPER volume 7 is divided into three parts Part I isdesigned more like a textbook by making a methodological and theoretical introduction,that will allow the reader to become familiarized with the image analysis jargon, and tofigure out what are the different steps required to obtain reliable results Image analysisimplies the following steps whatever the image application: image acquisition, calibrationand filtering (or pre-processing), image enhancement and classification (or processing),image analysis (or image interpretation) (Jongmans et al 2001) Part I tries to follow thislogical sequence In Chapter 2, Lamoureux and Bollmann review the different technologies(hardware) applicable for the study of lake and marine sediments, at a macroscopic andmicroscopic scale in order to obtain the best possible digital images Their contributionpoints out issues that must be considered to account for artifacts in the acquisition process,and prior to start the acquisition of an extensive set of images Chapter 3 by Nederbragt et al

describes software-based operations used to perform the analysis of images (sensu lato),

i.e., image calibration, image filtering and image classification, as well as how to transformpopular RGB files within the CIE L*a*b* systems, more useful for paleoenvironmentalreconstructions Pirard outlines in Chapter 4 the different kinds of measurements that can

be retrieved from images with a particular emphasis on the analysis of image intensities(gray levels, colors) and individual objects (size, shape, orientation) Pirard also discussthe problem of statistical representativity of the pixels and advocate for caution wheninterpreting the results In Chapter 5, Francus and Pirard illustrate how researchers cantest the validity of the results obtained using image analysis techniques, and advocate for

a systematic quality control of the results

4 FRANCUS, BRADLEY AND THUROW

Part II of the volume illustrates six applications of imaging techniques performed

on macroscopic (images of surface of sediment cores) and microscopic (slabs and sections) samples using miscellaneous supports (digital and analog photography, X-ray,electron microscopy) in order to reconstruct paleoenvironments Chapter 6, by Nederbragtand Thurow, outlines comprehensively how to extract color data from digital images ofsediment cores, focusing on techniques to filter out artifacts due to uneven illumination.Ortiz and O’Connell explain in Chapter 7 how to retrieve quantified information from oldernon-digital photographs, such as photographs of sediment cores from archived OPD andDSDP cruises In Chapter 8, Prokoph and Patterson describe an ingenious methodologyapplicable to annually laminated sediments that transforms digital sediment color data(recorded in a depth-scale) into a time-scale data set Chapter 9, by Principato, describes asimple methodology to quantitatively characterize diamictons from X-ray radiographs ofwhole or half sediment cores In Chapter 10, Ojala outlines how to acquire the best possibleX-radiographs of thin impregnated slabs of laminated sediments in order to perform thecounting and quantification of the laminae Then, Chapter 11, by Soreghan and Francus,reviews the issues during the acquisition of images using scanning electron microscopes

thin-in backscattered mode, and illustrates the analysis of ththin-in-sections of an old consolidatedloess deposit aiming for the reconstruction of paleowind intensity

The last Part outlines advanced techniques that may prefigure what the future of imageanalysis will be Bollmann et al describe in Chapter 12 robots that automatically acquireimages of microscopic samples (microfossils) aiming to process these images with auto-mated recognition systems, i.e., neural networks The following Chapter 13, by France et al.,focuses more on the software aspect of automated recognition by neural networks, providing

an example for automated recognition of pollen grains Finally, Verrecchia examples theuses of advanced mathematical tools, such as wavelet and multiresolution analysis in order

to analyze and retrieve measurements on images of banded/laminated samples To completethe book, a comprehensive glossary is included to help the reader to obtain a correctunderstanding of the words used through this somewhat technical volume

Computer scientists and engineers develop new powerful tools and algorithms every day.Geoscientists in general and sedimentologists in particular should take advantage of thesetechnological advances by looking for interdisciplinary collaborations Some accomplish-ments, such as the automated recognition of microfossils, are not fulfilled yet but are close tocompletion We need to better identify our needs in order to guide the next developments, andthis identification starts with a better understanding of what image analysis can accomplish.The future of image analysis techniques in paleoenvironmental science will probably bethe integration of processing algorithms within the acquisition phase, allowing the scientist

to concentrate on the analysis of the data sets produced (Jongmans et al 2001) The authorshope that this volume will trigger new ideas for the use of imaging techniques The topic

is new but the technique is very flexible, in such a way that “ imagination is the limit”(Saarinen and Petterson 2001)

AN INTRODUCTION TO IMAGE ANALYSIS . 5

data for paleoclimate reconstruction from marine and lacustrine sequences held at the

University of Massachusetts, Amherst, in November 2001 Pierre Francus is supported

by the University of Massachusetts, Amherst We thank Frank Keimig (Climate SystemResearch Center) for his help during the edition of this volume

Appendix: workshop recommendations

Proceeding with image analysis involves the same three major steps regardless of the type

of sample, e.g., surface of sediment core, thin-section, or the technique used to acquire animage (RGB photography, X-radiography, scanning electron microscopy) These steps areimage acquisition, image processing and image measurement

Image acquisition

It is emphasized that the quality of the image must be the best possible A lot of energyshould be spent on this step Acquiring images should involve:

Choice of the magnification, resolution, and size of image

One needs to consider the smallest feature that needs to be detected, the largest featurethat will be encountered and the representativity of the image with respect to the overallsample

Illumination

Variation of light intensity needs to be checked in the field of view, and the analyst must beaware of spatial and temporal variations To correct for irregular illumination, we recom-mend acquisition of a photograph of the background (for example 18% gray sheet) at thebeginning of the image acquisition session and at the end

Calibration standards

Where it is possible, spatial (ruler, grids) and color (gray/color chart, density wedges)references should be acquired on each photograph If not, the ruler and color/gray chartsshould be acquired at the beginning and the end of each working session, keeping in mindthe need to maintain the image acquisition conditions strictly constant during the acquisitionsession To maintain acquisition conditions strictly constant, it is also recommended thatimages should be acquired in the shortest period of time possible It will avoid miscellaneousproblems due to aging of color charts or filament, moving equipment to another location,changing hardware and software

Metadata

It is critical to record as much information as possible regarding the factors that can influencethe quality of images They include among other things, the characteristics and settings ofthe acquisition device (e.g., depth of field, current intensity in a SEM filament) and anyevent occurring during the working session such as a power failure The calendar of workingsession should also be noted

6 FRANCUS, BRADLEY AND THUROW

Image processing

In order to insure the intercomparability of the measurements it is necessary to documentthe software used and a detailed description of the filters used in the methods section ormetadata section of all published work It is also recommended to avoid software that isnot explicit in explaining algorithms used for processing For example, there are severalways to compute a perimeter The user needs to check what is the method used to do so, toensure comparability of different approaches

A digital master or archive version of the image should be saved for each capturedimage File format involving lossy compression, such as Joint Photographic Experts Group(JPEG), should be avoided by all means since compression involved loss of informationthat can not be recovered Uncompressed file formats, such as Tagged Image File Format(TIFF), are recommended

Image measurement

The representativity of the measurements made on digital images should always be kept

in mind because pixels are samples of an image, images are samples of the sample underinvestigation, the samples under investigation are a sample of the sediment of interest.Each step of the image analysis should be carefully tested using sets of calibrationimages or test images As a general principle, testing can be accomplished by slightlyvarying a single component of image acquisition condition or processing procedure —while maintaining the others strictly identical — and monitoring the impact on the finalmeasurements It is impossible to review all the tests that need to be conducted here because

of the variety of procedures However, the following steps should be carefully considered:

Related to image acquisition: magnification, resolution, contrast, brightness, color

coding systems (RGB, L*a*b*), hardware, image sampling representativity, illumination(spatial repartition, drift), spatial deformation (parallax effect), pixel shape, 8-bit <> 16-bitimages, TIFF <> other formats imposed by hardware and software, .

Related to image processing: noise removal filters, contrast/brightness manipulation,

image enhancement, segmentation and thresholding, edge detection, binary image ulation, .

manip-Related to image measurement: orientation, perimeters, distances, alignments, ellipse

fitting, and homemade indices

WEB site

The workshop attendees recommended the compilation of a web site where the followinginformation can be gathered:

List of references related to image analysis

List of hardware/software providers

Documentation of computer codes or filters used in research paper

A record and archive of metadata related to imaging techniques

A place to publish things that do not work

A place to publish testing of image analysis procedures

AN INTRODUCTION TO IMAGE ANALYSIS . 7There is a need for this because it is very difficult to publish such information in regularresearch papers The workshop participants agreed that such data are essential to insureintercomparison and reproducibility of results They also agreed this web page should bemaintained professionally.

Krumbein W.C 1941 Measurement and geological significance of shape and roundness ofsedimentary particles J Sed Petrol 11: 64–72

Russ J.C 1999 The Image Processing Handbook CRC Press, Boca Raton, Florida, 771 pp.Saarinen T and Petterson G 2001 Image analysis techniques In: Last W and Smol J (eds), TrackingEnvironmental Change Using Lake Sediments: Physical and Geochemical Methods KluwerAcademic Publishers, Dordrecht, The Netherlands, pp 23–39

Terribile F and Fitzpatrick E.A 1992 The application of multilayer digital image-processingtechniques to the description of soil thin-sections Geoderma 55: 159–174

VandenBygaart A.J and Protz R 1999 The representative elementary area (REA) in studies ofquantitative soil micromorphology Geoderma 89: 333–346

This page intentionally left blank

Part I: Getting started with Imaging Techniques

(or methodological introduction)

This page intentionally left blank

Department of Earth Sciences

ETH and University Zurich

Sonneggstrasse 5, 8092 Zurich

Switzerland

Keywords: Digital photography, Analog photography, Scanning, X-radiograph, Scanning electron microscope,

Color, Light filtering, Sedimentology, Image analysis, Paleoenvironmental reconstruction

Introduction

With increased interest in the use of sedimentary records for paleoenvironmental ysis, considerable effort has been made to utilize various image properties and analysistechniques as quantitative and semi-quantitative environmental proxies (Hughen et al.1996; Petterson et al 1999; Francus 1998; Nederbragt et al 2000; Nederbragt and Thurow2001; Tiljander et al 2002) For the most part, these approaches centre on the use ofimage information obtained from the sediments in the visible (400–750 nm) bands of theelectromagnetic spectrum Some researchers make use of near infrared and infrared (NIR,750–1200 nm), ultraviolet (UV, 1–400 nm) and X-ray regions of the electromagnetic spec-trum as well Increasingly, available technologies have extended these investigations intoimage analysis based on synthetic imagery produced from electron microscopy This type

anal-of imagery typically provides resolution anal-of features at the micron (µm) scale but may also

be used to study sediment properties at larger scales Therefore, significant improvements

in acquisition technologies, computing power and storage capacity have made sedimentaryimage processing increasingly viable for many applications in paleoenvironmental analysis

An essential first step in sedimentary image analysis research is the acquisition of highquality images that are suitable for the research objectives The diversity of available imageacquisition and processing systems reflects the varied interests and the resources available

to individual researchers Successful image acquisition requires substantial planning andconsideration of the inherent limitations of the selected acquisition method Clearly, poor

11

P Francus (ed.) 2004 Image Analysis, Sediments and Paleoenvironments Kluwer Academic

12 LAMOUREUX AND BOLLMANN

quality or substandard imagery will create significant problems during the subsequentanalysis and should be avoided where possible

This chapter is intended to provide an overview of image acquisition methods with phasis on the issues necessary to obtain high quality images required for quantitative imageanalysis Issues regarding the selection of a particular technique and major considerationsrelated to acquisition conditions are discussed, and are followed by brief descriptions ofthe common types of acquisition methods currently available for sedimentary analyses Fordetailed discussion of the analytical procedures used for extracting quantitative informationfrom sedimentary images (e.g., enhancement, calibration, and segmentation), the reader isreferred to the other chapters that follow in this volume

em-Image acquisition and paleoenvironmental research

It is tempting to begin using image analysis for a variety of paleoenvironmental researchwith relatively little consideration of the image acquisition process Indeed, a considerableamount of early work successfully made use of commonly available equipment to captureimages This apparent success has been largely in qualitative research, and limited tovisualization and archiving of sedimentary properties, perhaps with some enhancement

of image contrast or color However, quantitative image processing requires careful tion to a variety of conditions during the acquisition process (lighting, exposure) that arefrequently overlooked in qualitative analysis (Fig 1) Therefore, it is critical to establishoptimal acquisition conditions as a first step in any quantitative sedimentary image analysisproject Despite the differences in acquisition techniques, there are many common issues

atten-in obtaatten-inatten-ing high quality images

an image (Edmund Industrial Optics 2002) Resolution effectively determines the amount

of detail that may be obtained from the sample of interest Typically, the resolution of digitalhardware is reported in dots (pixels) per inch (dpi) or in pixel size (usually inµm) While

many inexpensive scanners and other acquisition hardware devices provide optical tions of 600–1200 dpi, many manufacturers of consumer products report higher resolutions

resolu-IMAGE ACQUISITION 13

Figure 1 An example of an image of laminated sediment from Sanagak Lake, Nunavut, Canada, obtained by

scanning an X-radiograph negative using a 600 dpi flatbed scanner (A) The image has a gradual shift to higher gray scale values (lighter) from right to left that was in part due to uneven exposure of the original X-radiograph film and also due to uneven acquisition by the scanner This image defect will lead to problems when assembling two adjacent and overlapping images (B) Plotted values from two separate X-radiograph scans from the same lake reveal a prominent downward trend with depth and demonstrate an offset in gray scale values where the two images overlap.

(9600+ dpi) that are produced from interpolation of the raw, optical scan With this type

of equipment, care should be taken to limit acquisition to the maximum optical resolution

of the hardware, to avoid uncontrolled interpolation by software drivers Where the opticalsystem can be adjusted with different lenses, the real resolution can be similarly adjusted.However, for a given camera, increased resolution will be at the expense of image scale,because the number of pixels in the camera sensor is fixed In cases where it is not clear whatthe resolution of the system is, or if the user wishes to test the effective resolution, speciallydesigned targets are available (Edmund Industrial Optics 2002) The most common used isthe United States Air Force (USAF) target, although other organisations (e.g., Institute ofElectrical and Electronics Engineers (IEEE)) produce similar tests

The appropriate combination of scale and resolution for image acquisition depends onthe subsequent analysis to be performed For visual purposes, including illustrations forpublication, lower resolutions (300–1200 dpi) are usually sufficient Stratigraphic observa-tions and measurements at the millimetre-scale can also be carried out at these resolutions

14 LAMOUREUX AND BOLLMANN

Figure 2 An example of the impact of acquisition resolution Sedimentological features and other qualitative

information in laminated sediments from Nicolay Lake, Nunavut, Canada are apparent from the 600 dpi scan in the left panel Enlargements (panels at right) of a small section containing isolated sand and silt grains shows the pixeling and degradation of sedimentological properties in the 600 dpi scan compared to the 2400 dpi scan The enlarged area is outlined on the lower magnification image.

(Fig 2) However, many quantitative studies (e.g., Francus (1998)) require substantiallyhigher image resolutions with pixel sizes<5 µm.

As a matter of practicality, the number of images required for a given project should beconsidered at an early stage While considerable progress has been made to automate orstreamline the time required to acquire images (Bollmann et al., this volume; Nederbragt

et al., this volume), many systems still require substantial time to obtain images After theimages are obtained, storage capacity may be a limiting factor, although this is increasinglybecoming less of an issue However, the human resources available for a project may be themain factor that limits the number of images that can be analysed The number of imagescan be controlled by the scale of images (for instance, a small number of images that eachcover a large section of sample) although the resulting changes to image resolution or scalemay hinder the analysis

Rapid developments in the hardware available for image capture have resulted in thewidespread use of direct digital acquisition of images from sedimentary materials Digitalcameras and scanners are the most commonly available to researchers because they arerelatively inexpensive and can be used for purposes other than sedimentary image analysis.While the trend towards improved features and sensor resolution is likely to continue,analog acquisition remains as a viable means to obtain high-quality images that can be sub-sequently digitized (e.g., Ortiz and O’Connell, this volume) Many microscopy systems areequipped with film cameras and provide excellent results Similarly, analog X-radiography

600 dpi

2400 dpi

5 mm

IMAGE ACQUISITION 15and electron microscopy systems are by far the most common equipment available toresearchers Therefore, while analog approaches require an additional digitizing step, theyremain important tools for many sedimentologists Like their digital counterparts, most ofthe key issues in image acquisition apply to analog equipment (see the next section).Finally, quantitative assessment of image properties and development of techniques thatcan be transported between laboratories require consideration of the reproducibility of theimage acquisition method Image properties of many sedimentary samples change rapidlywith age For example, reduced sediment will change composition due to oxidation Simi-larly, some image acquisition equipment (lights, film) also age with unpredictable results.Many of the key issues discussed in the following section are designed to document andminimize the influence of sample and equipment aging, as well as the varying acquisitionconditions Reproducible, quantitative results depend on accounting for these changes.

Key issues in image acquisition

Specific details of the image acquisition process will vary with the equipment used andthe sample material Regardless, the issues that affect image quality are frequently similar.This section discusses several of the most important considerations for obtaining the bestquality images, particularly sample shape, lighting and image calibration

Sample shape

Depending on the type of acquisition, the shape of the sedimentary sample will affect theresultant image substantially With sample shape, we refer to the spatial and geometric form

of the object under examination In principle, uniform geometry, particularly with respect

to the imaging sensor optical system is critical to obtain clear images In the broadest sense,shape affects focus with reflected light, such as camera or video systems For example, it will

be difficult to focus on a sediment slab that is uneven in thickness or has an irregular surface.Systems that utilize transmitted light (e.g., microscopes or scanners) are also sensitive toimage shape with respect to final image exposure For instance, a thin section of uneventhickness will appear as darker in the thickest areas compared to the thinnest areas Sampleshape is especially important for X-radiography as the attenuation of the X-ray beam iscontrolled, in part, by sample geometry (Principato, this volume)

Further, lens aberrations (or imperfections), particularly at low magnification can inducegeometric distortions in the final image Although this is typically a minor concern withhigh quality lens elements, this type of image distortion can be estimated by imaging auniform grid pattern to measure distortion across the field of view, if necessary

Image registration

Like sample shape, accurate image registration is necessary to ensure that the imageproperties can be indexed correctly to the true sample space Alignment of two or moreoverlapping images to produce a larger image requires the registration of the individualimages Misalignment can result in an offset between quantitative information from thesample stratigraphy and any related physical measurements In practice, registration can beaccomplished using one of several simple techniques, depending on the required accuracyand the type of imaging system In the case of large-scale core imaging, inclusion of a ruler

16 LAMOUREUX AND BOLLMANN

into the image is a common and useful registration method which permits subsequentmerging of information from overlapping samples Similarly, registration or measuringmarks are used on smaller samples to obtain correct image placement Where repeatedimages are taken of the same sample (e.g., thin sections), the use of alignment jigs canprovide accurate and consistent placement of the samples (Protz and VandenBygaart 1998)

Lighting considerations

The type and arrangement of the illumination source is an important consideration foracquisition using cameras and microscopes The selection of a particular type of lightingcan enhance the final image substantially, by improving contrast, reducing glare, or in somecases, providing an image that will require minimal or reduced post-processing By contrast,many acquisition systems, including scanners and X-radiography systems, are limited to asingle type of illumination Regardless, it is important to note that even fixed illuminationsources may vary as they age during use, thus changing the amount and wavelength of light

by high-wattage lights can heat fixtures rapidly and contact can cause burns

Most light sources used for photography or in scientific equipment have well mented illumination bands, and are frequently referred to by their black body tempera-ture For instance, commercial photography floodlights (typically 300–500 watt power)with tungsten filaments emit light at 3200◦K Although most illumination products are

docu-broad-band emitters, selective filtering of the light source can generate a relatively narrowbandwidth of illumination (see section below on filtering) It is important to note thatphotographic films that are designed for use with tungsten floodlights are available andwill provide better color rendition compared with conventional commercial films that areoptimized for general lighting sources

In addition to the different sources of light, the type of sample illumination will make asubstantial difference on the quality and properties in the final image Uneven illumination

or artefacts caused by inappropriate illumination are difficult to remove from images andmay prevent quantitative analysis Therefore, careful selection of the illumination canminimize or eliminate problems that are common with sedimentary materials, in particularglare from wet surfaces of sediment cores While the full range of illumination methodsare typically only available with photomicroscopy, some are readily used for larger-scalephotography (Fig 3)

Typically, most image problems are associated with highly directional lighting, duced for example, by photographing a sample with two bright spotlights In addition tothe uneven illumination, directional lighting typically produces glare and shadows Whilethese problems can be alleviated to some degree by using multiple light sources, directionallighting is rarely satisfactory for sediment imaging By contrast, diffuse illumination, pro-duced by a large number of lights or by filtering can minimize glare and shadows Diffuse

pro-IMAGE ACQUISITION 17

Figure 3 A schematic representation of different illumination strategies for imaging sedimentary samples.

Although the figure shows different lighting configurations for photomicroscopy, many are suitable for other acquisition techniques Note that directional and external diffuse light usually come from more than one source, but have been simplified in the figure.

light sources are typically obtained by reflecting the light source on a diffusion screen, atechnique commonly used in commercial photography For photomicroscopy, diffuse lightsources are available for both reflected and transmitted light imaging In the former case, alarge number of small light sources (typically light emitting diodes (LED)) are configured

in a ring around the lens and provide coaxial light (Edmund Industrial Optics 2002) Incases where glare or shadowing remain problematic, diffuse axial illumination producedusing a beamsplitter in the microscope lens assembly generates very even diffuse light withminimal glare

Where transmitted light imaging is desired, a diffuse backlighting source is of criticalconcern Diffusing lenses and sample stages are available to provide even, diffuse light forimaging transparent and semi-transparent samples (e.g., thin sections) Where the samplematerial is opaque, brightfield (or backlight) illumination can provide images that elimi-nate surface details and have high edge contrast This illumination technique is especiallyuseful for counting objects such as sediment grains or charcoal and has found wide use insedimentology (Bouma 1969)

The selection of an appropriate illumination configuration and source will frequentlyprovide images with relatively uniform lighting While many light sources appear to providereasonably even, diffuse light during sample inspection and visual assessment of the images,quantitative image analysis requires estimation of the uniformity of the illumination Bydoing so, uneven illumination can be identified and corrected (see Nederbragt et al (thisvolume)) A pragmatic approach requires obtaining a control image without a sedimentary

DiffusionScreen

Stage

18 LAMOUREUX AND BOLLMANN

sample Analysis of the control image will identify areas of reduced illumination that areinherent in the acquisition system However, it is important to note that this technique isnot suitable for accounting for uneven illumination caused by glare or shadows, as theseartefacts are specific to each sample Similarly, analysis of the control image can provide animportant measure of the noise generated by the image capturing system In theory, perfectacquisition of a uniformly colored object using a uniform light source should generate animage with identical sensor values throughout the image The deviation from this idealsituation is an important measure of the noise introduced during the sampling process.Finally, while it is not usually possible to predict how the illumination source will age,care should be taken to provide a means to measure its impact on a series of images Again, apragmatic and simple approach to this problem is to include one or more color standards ineach image In this way, changing illumination properties can be quantified (see Nederbragt

et al (this volume))

Light filtering

Light filtering can be used for a variety of purposes in image acquisition to improvethe image characteristics for subsequent processing By making use of filter properties,the reflection (and transmission) properties of the sedimentary material can be enhanced

to improve contrast between different components or structures Additionally, filters aretypically necessary to make use of infrared and ultraviolet wavelengths In general, filtersused in image acquisition can be categorized into bandpass, polarizing, and neutral densityfilters Each type of filter has specific properties that affect the spectrum of light or thequantity of light that passes through the filter The properties and uses of the main filtertypes are discussed below

Bandpass filters are all characterized by reduction of light transmission in a particularrange of the electromagnetic spectrum This group of filters can be further divided into sev-eral important functional types, including: interference, bandpass, longpass and shortpasstypes Specifically, shorter wavelengths correspond to ultraviolet and blue light while longerwavelengths produce red and infrared light Bandpass filters are designed to transmit lightonly in a specified range of wavelengths and to absorb wavelengths outside of this range(Fig 4) In practice, the transmission bandwidth is bounded by transition bands of partialtransmission; therefore, most bandwidth filters do not produce an ideal filtering effect.Interference filters are special case of bandpass filters that have very narrow absorptionbands (±10 nm) These filters are used in a variety of spectroscopic analytical equipmentand are available in a large range of central wavelength values Francus and Pirard (thisvolume, Fig 3) show an example of how such a filter help to discriminate between twodifferent mineral phases However, unlike more conventional bandpass filters, interferencefilters are very sensitive to placement and variations in the incident angle of illuminationcan impact the filtration characteristics (Edmund Industrial Optics 2002) Longpass andshortpass filters are essentially bandpass filters that absorb on the upper and lower ranges

of the imaging spectrum

Depending on the specified wavelength absorption characteristics of the filter in use,the illumination on the sediment sample can be varied substantially Moreover, filters can

be used in combination to further alter the illumination Indeed, this principle is used incolor photographic film where the incident light on the film is selectively absorbed bylight sensitive materials, separated by thin film filters that restrict the passage of different

IMAGE ACQUISITION 19

Figure 4 Generalized wavelength filtering characteristics of common colored filters Additive colored filters (A)

allow passage of the corresponding band of light (e.g., a green filter allows the passage of only green light), while subtractive filters (B) block the transmission of one of the additive primaries (e.g., magenta blocks green light) The gray shaded area is the transition zone for the red filter.

wavelengths Color filters are available as dichroic (surface coated) or solid substrate typesand are usually available as sets using either additive or subtractive color systems Additiveprimary color filters (red, green, and blue) transmit light in their respective color, andwhen all three are combined together, produce white light (Lillesand and Kieffer 1984)(Fig 4) In contrast, the subtractive primary colors (yellow, magenta, and cyan) filter

by removing one of the primary additive colors For instance, filtering out green lightresults in transmission of blue and red light with the net result of magenta While usedfor different processes, additive filters are used primarily in projection applications liketelevision and subtractive filters are used in color photographic film Both types of filtrationhave application in sedimentary image analysis through enhancement of contrast betweencolors and for identifying mineral materials (e.g., Protz and VandenBygaart (1998)) It

is important to note that color filtering can also be carried out during subsequent imageanalysis (see Nederbragt et al (this volume))

Polarizing filters are most frequently used to minimize glare from wet sediment surfacesand cross polarization can be used for some mineral identification (Fig 5) Ordinary light(random polarization) can be thought of as consisting of electromagnetic waves at rightangles to one another, and occurring in all possible aspects from the light source (Ed-mund Industrial Optics (2002)) Polarizing filters reduce the light to a single plane, hence

20 LAMOUREUX AND BOLLMANN

Figure 5 An example of the use of cross-polarization to distinguish sedimentary components Laminated

sed-iments (9–12 cm below the lake floor) from the deepest point in Baldeggersee, Switzerland contain carbonate lamina produced during diatom blooms These layers are apparent in an image obtained from a thin section (a) Cross-polarization of the same thin section during scanning (b) improves the clarity of the carbonate layers (homogeneous, light gray lamina) for subsequent analysis Both images were obtained using an Agfa Duoscan transparency scanner at 1440 dpi in 8-bit RGB mode The images have been converted to 8-bit greyscale and the brightness and contrast enhanced for presentation (images courtesy of P Francus).

substantially reducing the amount of light transmitted If a second polarizing element isadded, the amount of light transmitted is proportional to the relative transmission axes of thetwo filters If the polarizers are perpendicular, no light is transmitted If the two polarizers areparallel to each other, the transmission is considerably higher Many microscopes have fully

1 cm

IMAGE ACQUISITION 21integrated cross-polarizing filter sets built into the transmitted light stage for thin sectionwork (Fig 5) Alternatively, polarizing filters can be added to camera and microscopelenses, and sheet film filters can be placed on flatbed scanners with sample material toobtain similar effects.

Finally, neutral density (or gray) filters are designed to evenly reduce the transmission

of all wavelengths of light to prevent overexposure or damage to the imaging equipment Inpractice, the use of neutral density filters is uncommon in most image acquisition situations,but may be used to protect lens surfaces and sensitive sensor systems where necessary

Calibration of image density

As discussed in the illumination section, calibration of the image density in photographyrequires the inclusion of a density standard composed of a color or grayscale pattern target

in each image Grayscale patterns are usually composed of a series of rectangles varyingfrom white to black, typically in 15 or more density steps Color calibration requires morespecialized targets (see Nederbragt et al (this volume)) Therefore, before acquisitionbegins, it is important to develop the image analysis strategy and protocols to select anappropriate calibration standard Finally, image calibration targets also serve to indicatethe impact of illumination aging, thus serving as an important quality control measure

Sample preparation for image acquisition

A comprehensive discussion of sample preparation for image acquisition is beyond thescope of this chapter Detailed descriptions of the preparation procedures can be found inthe accompanying chapters of this volume However, as many sampling procedures (forimaging and other analyses) are destructive, it is important to consider the type of imageacquisition prior to extensive work with the sediment (Fig 6) Many image acquisitionmethods require minimal preparation work, including whole- and half-core X-radiographyand core face photography Protocols have been developed to consistently prepare core facesfor acquisition for color analysis (see Nederbragt and Thurow (this volume)) and resultsfrom these studies have produced valuable paleoenvironmental proxy records (Hughen et al.1996; Nederbragt et al 2000; Nederbragt and Thurow 2001) Generally, clear core faceimages can be obtained by carefully scraping the surface, moving the cleaning edge acrossthe sedimentary structures In cases where the sediment is clay-rich, an electro-osmoticknife can be used to prevent smearing (Pike and Kemp 1996) If the sediment is frozen,covering the surface with a thin film of water can be useful to reduce reflections off icecrystals (Renberg 1981) Finally, in some cases, particularly with some clastic sediments,allowing the surface to partially dry improves the contrast between sedimentary structures,although this may lead to cracking, particularly in diamictons

Image acquisition at finer scales usually requires some degree of subsampling In thecase of unconsolidated materials, stabilization of the sediment is also usually required.Sediments can be stabilized in blocks and slabs using a variety of techniques (see Bouma(1969)) depending on the sample texture and water content The most difficult case is clay-rich, wet sediments that require dehydration using either freeze-drying or liquid-liquidmethod, followed by impregnation with low viscosity epoxy resin (Pike and Kemp 1996;Lamoureux 2001) The stabilized sediment samples are suitable for a variety of image