The Image Processing Handbook
Third Edition by John C. Russ - CRC Press
Table of Contents
Introduction
Acknowledgments
Chapter 1—Acquiring Images
Human reliance on images for information
Using video cameras to acquire images
Electronics and bandwidth limitations
High resolution imaging
Color imaging
Digital cameras
Color spaces
Color displays Image types
Range imaging
Multiple images
Stereoscopy
Imaging requirements
Chapter 2—Printing and Storage
Printing Dots on paper
Color printing Printing hardware
Film recorders
File storage
Optical storage media
Magnetic recording
Databases for images
Browsing and thumbnails
Lossless coding
Color palettes
Lossy compression
Other compression methods
Digital movies
Chapter 3—Correcting Imaging Defects
Noisy images
Neighborhood averaging
Neighborhood ranking
Other neighborhood noise reduction methods
Maximum entropy
Contrast expansion
Nonuniform illumination
Fitting a background function
Rank leveling
Color shading
Nonplanar views
Computer graphics
Geometrical distortion
Alignment
Morphing
Chapter 4—Image Enhancement
Contrast manipulation
Histogram equalization
Laplacian Derivatives
The Sobel and Kirsch operators
Rank operations
Texture Fractal analysis
Implementation notes
Image math
Subtracting images
Multiplication and division
Chapter 5—Processing Images in Frequency Space
Some necessary mathematical preliminaries
What frequency space is all about
The Fourier transform Fourier transforms of real functions
Frequencies and orientations
Measuring images in the frequency domain
Orientation and spacing
Preferred orientation
Texture and fractals Filtering images Isolating periodic noise
Masks and filters
Selection of periodic information
Convolution and correlation
Fundamentals of convolution Imaging system characteristics
Removing motion blur and other defects
Template matching and correlation
Autocorrelation
Conclusion
Chapter 6—Segmentation and Thresholding Thresholding
Multiband images
Two- Dimensional thresholds
Multiband thresholding
Thresholding from texture
Multiple thresholding criteria
Textural orientation
Accuracy and reproducibility
Including position information
Selective histograms
Boundary lines
Contours Image representation
Other segmentation methods
The general classification problem
Chapter 7—Processing Binary Images
Boolean operations
Combining Boolean operations
Masks From pixels to features
Boolean logic with features
Selecting features by location
Double thresholding
Erosion and dilation
Opening and closing Isotropy
Measurements using erosion and dilation
Extension to grey scale images
Coefficient and depth parameters
Examples of use The custer
Skeletonization Boundary lines and thickening
Euclidean distance map
Watershed segmentation
Ultimate eroded points
Fractal dimension measurement
Medial axis transform
Cluster analysis
Chapter 8—Image Measurements
Brightness measurements
Determining location
Orientation Neighbor relationships
Alignment Counting features
Special counting procedures
Feature size
Caliper dimensions
Perimeter Ellipse fitting Describing shape
Fractal dimension
Harmonic analysis
Topology Feature identification
Three- Dimensional measurements
Chapter 9—3D Image Acquisition
Volume imaging vs. sections
Basics of reconstruction
Algebraic reconstruction methods
Maximum entropy
Defects in reconstructed images
Imaging geometries
Three dimensional tomography
High resolution tomography
Chapter 10—3D Image Visualization
Sources of 3D data
Serial sections Optical sectioning
Sequential removal
Stereo 3D data sets
Slicing the data set
Arbitrary section planes
The use of color
Volumetric display
Stereo Viewing Special display hardware
Ray tracing
Reflection Surfaces Multiply connected surfaces
Image processing in 3D Measurements on 3D images
Conclusion
Chapter 11—Imaging Surfaces
Producing surfaces
Devices that image surfaces by physical contact
Noncontacting measurements
Microscopy of surfaces
Surface composition imaging
Processing of range images
Processing of composition maps
Data presentation and visualization
Render- Ing and visualization
Analysis of surface data
Profile measurements
The Birmingham measurement suite
New approaches topographic analysis and fractal dimensions
References
Index
Introduction
Image processing is used for two somewhat differ- Ent purposes:
a) Improving the visual appearance of images to a human viewer, and
b) Preparing images for measurement of the features and structures present.
The
techniques that are appropriate for each of these tasks are not always
the same, but there is considerable overlap. This book covers methods
that are used for both purposes. To do the best possible job, it helps
to know about the uses to which the processed images will be put. For
visual enhancement, this means having some familiarity with the human
visual process, and an appreciation of what cues the viewer responds to
in images.
It
also is useful to know about the printing process, since many images
are processed in the context of reproduction or transmission. The
measurement of images generally requires that features be well defined,
either by edges or unique (and uniform) Brightness or color, texture, or
some combination of these factors. The types of measurements that will
be performed on entire scenes or individual features are important in
determining the appropriate processing steps. It may help to recall that
image processing, like food processing or word processing, does not
reduce the amount of data present but simply rearranges it.
Some
arrangements may be more appealing to the senses, and some may convey
more meaning, but these two criteria may not be identical nor use
identical methods. This handbook presents an extensiive collection of
image processing tools, so that the user of computerbased systems can
both understand those methods provided in packaged software, and program
those additions which may be needed for particular applications.
Comparisons
are presented of different algorithms that may be used for similar
purposes, using a selection of representative pictures from light and
electron microscopes, as well as macroscopic, satellite and astronomical
images. In revising the book for this new edition, I have tried to
respond to some of the comments and requests of readers and reviewers.
New chapters on the measurement of images and the subsequent
interpretation of the data were added in the second edition, and now
there is a major new section on the important subject of surface images
which includes both processing and measurement.
The
sections on the everadvancing hardware for image capture and printing
have been expanded and information added on the newest technologies.
More examples have been added in every chapter, and the reference list
expanded and brought up to date. However, I have resisted suggestions to
put “more of the math” into the book. There are excellent texts on
image processing, compression, mathematical morphology, etc. , that
provide as much rigor and as many derivations as may be needed.
Many
of them are referenced here. But the thrust of this book remains
teaching by example. Few people learn the principles of image processing
from the equations. Just as we use images to “do science,” so most of
us use images to learn about many things, including imaging itself. The
hope is that by seeing what various operations do to representative
images, you will discover how and why to use them. Then, if you need to
look up the mathematical foundations, they will be easier to understand.
The
reader is encouraged to use this book in concert with a real source of
images and a computer- Based system, and to freely experiment with
different methods to determine which are most appropriate for his or her
particular needs.
Selection
of image processing tools to explore images when you don't know the
contents beforehand is a much more difficult task than using tools to
make it easier for another viewer or a measurement program to see the
same things you have discovered. It places greater demand on computing
speed and the interactive nature of the interface. But it particularly
requires that you become a very analyt- Ical observer of images.
If
you can learn to see what the computer sees, you will become a better
viewer and obtain the best possible images, suitable for further
processing. To facilitate this hands- On learning process, I have
collaborated with my son, Chris Russ, to produce a CD- ROM that can be
used as a companion to this book. The Image Processing Tool Kit contains
more than 200 images, many of them the examples from this book, plus
over 100 Photoshop- Compatible plugins that implement many of the
algorithms discussed here. These can be used with Adobe Photoshop® or
any of the numerous programs (some of them free) That implement the
Photoshop plugin interface, on either Macintosh or Windows computers.
Many
of the images were acquired directly from various microscopes and other
sources using color or monochrome video cameras and digitized directly
into the computer. Others were digitized using a digital camera
(Polaroid DMC), and some were obtained using a 24- Bit color scanner,
often from images supplied by many coworkers and researchers.
These
are acknowledged wherever the origin of an image could be determined. A
few examples, taken from the literature, are individually referenced.
The book was produced by directly making colorseparated films with an
imagesetter without intermediate hard copy, negatives or prints of the
images, etc.
Amongst
other things, this means that the author must bear full responsibility
for any errors, since no traditional typesetting was involved. (It has
also forced me to learn more than I ever hoped to know about some
aspects of this technology!) However, going directly from disk file to
print also shortens the time needed in production and helps to keep
costs down, while preserving the full quality of the images.
Special
thanks are due to Chris Russ (Reindeer Games, Inc. , Asheville, NC) Who
has helped to program many of these algorithms and contributed
invaluable com- Ments, and to Helen Adams, who has proofread many pages,
endured many discussions, and provided the moral support that makes
writing projects such as this possible.
Chapter 1 Acquiring Images
Human
reliance on images for information Humans are primarily visual
creatures. Not all animals depend on their eyes, as we do, for 99% or
more of the information received about their surroundings. Bats use high
frequency sound, cats have poor vision but a rich sense of smell,
snakes locate prey by heat emission, and fish have organs that sense
(and in some cases generate) Electrical fields. Even birds, which are
highly visual, do not have our eye configuration.
Their
eyes are on opposite sides of their heads, providing nearly 360- Degree
coverage but little in the way of stereopsis, and they have four or
five different color receptors (we have three, loosely described as red,
green, and blue). It is difficult to imagine what the world “looks
like” to such animals.
Even
the word “imagine” contains within it our bias towards images, as does
much of our language. People with vision defects wear glasses because of
their dependence on this sense. We tolerate considerable hearing loss
before resorting to a hearing aid and there are, practically speaking,
no prosthetic devices for the other senses. This bias in everyday life
extends to how we pursue more technical goals as well. Scientific
instruments commonly produce images to communicate results to the
operator, rather than generating an audible tone or emitting a smell.
Space
missions to other planets and Comet Halley always include cameras as
major components, and we judge the success of those missions by the
quality of the images returned.
This
suggests a few of the ways in which we have extended the range of our
natural vision. Simple optical devices such as microscopes and
telescopes allow us to see things that are vastly smaller or larger than
we could otherwise. Beyond the visible portion of the electromagnetic
spectrum (a narrow range of wave- Lengths between about 400 and 700
nanometers) We now have sensors capable of detecting infrared and
ultraviolet light, X- Rays, and radio waves. Figure 1 shows an example,
an image presenting radio telescope data in the form of an image in
which grey scale brightness represents radio intensity.
Such devices and presentations are used to further extend our imaging capability.
Figure
1 Radio astronomy produces images such as this view of NGC1265. These
are offen displayed with false colors to emphasize subtle variations in
signal brightness. Signals other than electromagnetic radiation can be
used to produce images as well. Acoustic waves at low frequency produce
sonar images, while at gigahertz frequencies the acoustic microscope
produces images with resolution similar to that of the light microscope,
but with image contrast that is produced by local variations in the
attenuation and refraction of sound waves rather than light. Figure 2
shows an acoustic microscope image of a composite material. Figure 2
Scanning acoustic microscope image (with superimposed signal profile
along one scan line) Of a polished cross section through a composite.
The central white feature is a fiber intersecting the surface at an
angle. The arcs on either side are interference patterns which can be
used to measure the fiber angle.
Figure
3 An electron diffraction pattern from a thin foil of gold. The ring
diameters correspond to diffraction angles that identify the spacings of
planes of atoms in the crystal structure. Figure 4 A convergent beam
electron diffraction (CBED) Pattern from an oxide microcrystal, which
can be indexed and measured to provide high accuracy values for the
atomic unit cell dimensions.
Figure
5 Typical graphics used to communicate news information (Source: USA
Today). Some images such as holograms or electron diffraction patterns
are recorded in terms of brightness as a function of position, but are
unfamiliar to the observer. Figures 3 and 4 show electron diffraction
patterns from a transmission electron microscope, in which the atomic
structure of the samples is revealed (but only to those who know how to
interpret the image). Other kinds of data including weather maps with
isotherms, elaborate graphs of business profit and expenses, and charts
with axes representing time, family income, cholesterol level, or even
more obscure parameters, have become part of our daily lives.
Figure
5 shows a few examples. The latest developments in computer interfaces
and dis- Plays make extensive use of graphics, again to take advantage
of the large bandwidth of the human visual pathway. There are some
important differences between human vision, the kind of information it
yields from images, and the ways in which it seems to do so, compared to
the use of imaging devices with computers for technical purposes. Human
vision is primarily qualitative and comparative, rather than
quantitative. We judge the relative size and shape of objects by
mentally rotating them to the same orientation, overlapping them in
ourminds, and performing a direct comparison.
This
has been shown by tests in which the time required to recognize
features as being the same or different is proportional to the degree of
misorientation or intervening distance.
Figure
6 shows an example. Figure 6 Several views of a complex three-
Dimensional figure. Which of the representations is/ Are identical and
which are mirror images? The time needed to decide is proportional to
the misalignment, indicating that we literally “turn the objects over”
in our minds to compare them….
Nhận xét
Đăng nhận xét