Sampled Imaging Systems: Analysis and Evaluation of
Sampled Imaging Systems: Analysis and Evaluation of
Sampled Imaging Systems: Analysis and Evaluation of
Sampled
Imaging
Systems
Sampled
Imaging
Systems
Richard H. Vollmerhausen
Donald A. Reago, Jr.
Ronald G. Driggers
Vollmerhausen, Richard H.
Analysis and evaluation of sampled imaging systems / Richard H. Vollmerhausen,
Donald Reago, and Ronald G. Driggers.
p. cm. -- (Tutorial texts in optical engineering ; v. TT87)
Includes bibliographical references and index.
ISBN 978-0-8194-8077-4 (alk. paper)
1. Imaging systems--Image quality. 2. Image processing--Statistical methods.
3. Fourier analysis. 4. Sampling (Statistics) I. Reago, Donald. II. Driggers, Ronald
G. III. Title.
TK8315.V6495 2010
621.36'7--dc22
2009053629
Published by
SPIE
P.O. Box 10
Bellingham, Washington 98227-0010 USA
Phone: +1 360.676.3290
Fax: +1 360.647.1445
Email: [email protected]
Web: http://spie.org
All rights reserved. No part of this publication may be reproduced or distributed in any
form or by any means without written permission of the publisher.
The content of this book and CD-ROM reflects the work and thought of the author(s).
Every effort has been made to publish reliable and accurate information herein, but the
publisher is not responsible for the validity of the information or for any outcomes
resulting from reliance thereon.
The authors make no warranties that the software programs contained on the enclosed
CD-ROM are free of error, or are consistent with any standard of merchantability, or will
meet the requirements for any particular application. CD-ROM Copyright © 2010 RHV
Consulting, LLC, 760 Jacktown Rd., Lexington, VA 24450.
James A. Harrington
Rutgers University
This tutorial is written for the design engineer or system analyst interested in
quantifying the performance of electro-optical imagers. Advancing technology in
detector arrays, flat panel displays, and digital image processing provide new
opportunities to expand imaging applications and enhance system performance.
Technical managers and design engineers are faced with evaluating the cost,
weight, and performance of an ever-expanding selection of technology options.
This book provides the theory and procedures for performance assessment.
This text supersedes Analysis of Sampled Imaging Systems, which was
published by SPIE Press in 2000. Part I updates the earlier work. Part II discusses
performance evaluation of electro-optical imagers. Part III provides computer
programs and up-to-date information on detector arrays, optics, and display
options. This book provides the theory, procedures, and information needed to
evaluate and compare the performance of available imaging technologies.
Our prior work Analysis of Sampled Imaging Systems focused on the
mathematical formulism needed to analyze sampled imagery. That book
described the sampled imager response (SIR) function. SIR quantified sampled
imager aliasing as well as the system transfer response. Fourier transform theory
was used to describe and quantify sampling artifacts such as display raster,
blocky images, and the loss or alteration of image detail due to aliasing.
However, the metrics provided by the earlier book were “rules of thumb.”
Sampled imager design rules were based on experience and experimentation. No
theory existed to quantify the effect of aliasing on visual task performance. The
earlier work provided guidance on how to optimize sampled imagers by
minimizing aliasing. Analysis of Sampled Imaging Systems did not provide a
procedure to quantify the impact of aliasing on performance.
In the intervening years since the first book, we have discovered that the
effect of aliasing on targeting performance is predictable by treating aliasing as
noise. This book presents a resolution metric that predicts the effect of imager
blur, noise, and sampling on the probability of correctly identifying targets. This
new publication includes quantitative procedures for evaluating target acqui-
sition performance.
Part I of this book includes all of the pertinent material from Analysis of
Sampled Imaging Systems. The first five chapters remain substantially the same
as the previous work. These chapters introduce sampling concepts and describe
the differences between shift-variant and shift-invariant systems. Chapter 2 on
Fourier optics is extensively rewritten. The errors associated with assuming
xi
Richard H. Vollmerhausen
Donald A. Reago, Jr.
Ronald G. Driggers
February 2010
xiii
images the scene onto a staring detector focal plane array (FPA). Diffraction and
aberrations blur the image. In the figure, the image on the FPA is upside-down as
well as blurred because the optical system inverts the image.
The detector array consists of a number of rows and columns of individual
detectors, as shown in the inset in Fig. 1.2. Photodetection occurs over the active
area of these individual detectors. The incoming light generates photo-electrons.
Since each detector integrates photo-signal over a finite active area, the detector
itself also blurs the image. Individual points in the optical image are summed
together if they fall on the same detector, and this summing of adjacent points
causes a blur.
The detector photocurrent is integrated (in a capacitor, charge well, or by
some other mechanism) for a period of time. Periodically (generally, every
sixtieth of a second in the U.S. or every fiftieth of a second in Europe), the
resulting signal charge is read out, and the integrator is reset. The amount of
charge from each detector depends directly on the intensity of light falling on that
detector. The charge output from each detector represents a sample of the lens-
and detector-blurred scene intensity. Note that, in the example shown in Fig. 1.2,
the active detector area does not cover the entire FPA. This array does not have a
100% fill factor.
scene
camera
lay
disp
Figure 1.1 Observer viewing the display of a sampled image.
Figure 1.2 Components of a camera. The lens images a scene onto the detector array.
Optical diffraction and aberrations in the lens blur the image. The detector further blurs the
image by integrating signal over the active detector area. Each detector provides one
sample of the blurred scene.
The two images in Fig. 1.3 summarize the action of the camera. Concept-
ually, the optical and detector blurs are lumped together and called the presample
blur. The image with optical and detector presample blur applied is shown in the
left-hand image in Fig. 1.3. The detectors then convert the light intensity at
specific locations in the blurred image to electrical signals. The electrical signals
represent image samples. In the right-hand image, the white dots indicate the
locations where the blurred image is sampled by the detector array.
A display device is used to reconstruct the image from the detector
(electrical) samples. The display device consists of an array of individual display
pixels. A “display pixel” is an individual light-emitting area on the display
surface. In the simplest case, the number and location of pixels in the display
correspond to the number and location of detectors in the FPA. The brightness of
each display pixel is proportional to the photo-signal from the corresponding
detector.
Figure 1.4 shows the display. An individual display pixel is shown in the
upper left-hand corner of the image. These pixels happen to be square. In this
example, there is one display pixel for each sensor sample shown in Fig. 1.3. The
intensity of each pixel shown in Fig. 1.4 is proportional to the photo-intensity at
the corresponding sample location in Fig. 1.3.
Figure 1.3 Left-hand image is blurred by the optics and the detector. Right-hand image
shows the location of detector samples as white dots.
Figure 1.4 Display of a sampled image. An individual pixel is shown in the upper left-hand
corner. Each pixel on the display is illuminated in proportion to the amplitude of the
sample from the corresponding location in Fig. 1.3.
(a) (b)
(c) (d)
Figure 1.5 (a) Original picture. (b) Presample blur applied to original picture. (c) Picture
showing locations of samples. (d) Picture showing reconstructed image with a single
display pixel shown in the upper left-hand corner.
All three stages of the sampling process are necessary: preblur of the image,
sampling, and postblur or reconstruction. Figure 1.6(a) shows the result of
sampling the original image without prefiltering [that is, Fig. 1.5(a) is sampled
rather than Fig. 1.5(b)]. In this case, aliasing hurts the final, displayed image.
Figure 1.5(d) looks more like Fig. 1.5(a) than Fig. 1.6(a) does.
Figure 1.6(b) shows the image samples displayed as points rather than as the
large pixels used in Fig. 1.5(d). In Fig. 1.6(b), the image is not blurred by the
display pixel, and the image cannot be integrated by the eye. To get a good
image, display reconstruction using a postsample blur is necessary.
As an illustration, move Fig. 1.6(b) close to the eye so that only points are
seen. Now, slowly move the figure away from the eye. Lena begins to appear as
the figure moves away because eye blur acts as a reconstruction filter.
Rules can be established for determining the optimum relationship between
preblur, sample spacing, and postblur. A well-sampled imaging system is one in
which the spacing (in milliradians or millimeters) between image samples is
small compared to the width of the presample blur. In this case, sampling
artifacts are not apparent in the image.
(a) (b)
Figure 1.6 The picture in (a) shows the original image in Fig. 1.5(a) sampled without pre-
blur and then reconstructed in a similar manner to Fig. 1.5(d). Figure 1.5(d) looks more
like Fig. 1.5(a) than Fig. 1.6(a) does because some presample blurring is necessary. The
picture in (b) is constructed with pixels that are spatially much smaller than the pixel pitch.
The pixels in (b) do not blur the image, and the picture is very hard for the eye to integrate.
Postblur of the image samples is also necessary.
Figure 1.7 Picture of Lena reconstructed using the same samples as were used in Fig.
1.5(d). This picture is improved because the display pixel shape provides a better match
to the original image between sample points.