Beruflich Dokumente
Kultur Dokumente
Computer Vision
Algorithms in
Image Algebra
second edition
H A N D B O OK OF
Computer Vision
Algorithms in
Image Algebra
second edition
Gerhard X. Ritter
Joseph N. Wilson
CRC Press
Boca Raton London New York Washington, D.C.
disclaimer Page 1 Monday, August 21, 2000 2:37 PM
Ritter, G. X.
Handbook of computer vision algorithms in image algebra / Gerhard X. Ritter, Joseph
N. Wilson.--2nd ed.
p. cm.
Includes bibliographical references and index.
ISBN 0-8493-0075-4 (alk. paper)
1. Computer vision--Mathematics. 2. Image processing--Mathematics. 3. Computer
algorithms. I. Wilson, Joseph N. II. Title.
This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with
permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts have been made to publish
reliable data and information, but the author and the publisher cannot assume responsibility for the validity of all materials
or for the consequences of their use.
Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical,
including photocopying, microfilming, and recording, or by any information storage or retrieval system, without prior
permission in writing from the publisher.
The consent of CRC Press LLC does not extend to copying for general distribution, for promotion, for creating new works,
or for resale. Specific permission must be obtained in writing from CRC Press LLC for such copying.
Direct all inquiries to CRC Press LLC, 2000 N.W. Corporate Blvd., Boca Raton, Florida 33431.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation, without intent to infringe.
1 The iac++ library supports the use of image algebra in the C++ programming language and is available
via anonymous ftp from ftp://ftp.cise.ufl.edu/pub/src/ia/.
Symbol Explanation
Uppercase characters represent arbitrary sets.
Lowercase characters represent elements of an arbitrary set.
Bold, uppercase characters are used to represent point sets.
Bold, lowercase characters are used to represent points, i.e.,
elements of point sets.
The set
!"!#! $ .
% %'& %)( The set of integers, positive integers, and negative integers,
respectively.
%+* The set
%* , - ""!#!!#.0/1"$ .
%'*& The set
%'*& , "!#!!#.+$ .
%)2 * The set
%'2 * 3 /4.6578#!"!"!
/98 #8#!!#!:.;/<=$ .
> >?& >@( >BADC The set of real numbers, positive real numbers, negative real
numbers, and positive real numbers including 0, respectively.
E The set of complex numbers.
F An arbitrary set of values.
F?G The set
F H $ .
unioned with
F(G F
The set unioned with
/ H $.
F?2 G F
The set unioned with
/ H H $.
I The empty set (the set that has no elements).
=J The power set of (the set of all subsets of ).
K "is an element of."
KL "is not an element of."
M "is a subset of."
N W
N _
`8a-b ` Let c . ` N ` `8a-V3b Wbed[Y8a d[family
Z[N ` ofe \^sets opZ c _ . by an indexing set
g=iliindexed
`8a-b
q N r q N VwN UN x y#y"y
0N
rts)u rvs)u r u q.
z N r z N r V3W#d{YdUZ[N r e \^?g=ili~Z'?_ .
rts)u rvs)u
N[Q
N[QPVW8d):8YdUZ{N{6Z[Q0_ .
Cartesian product
Symbol Explanation
©«ªR¬ If © ¬ pZ q , then ©0ª®¬ ,
V d u ª u )y#y"y§d q ª q .
©[¯¬ If © ¬
pZ q , then ©U¯¬ , V d u ¯ u )y#y"y§d q ¯ q .
© 2001 by CRC Press LLC
Symbol Explanation
°{±² If °?³²´pµ¶ , then °[±
²{·3¸¹+º¼»8º-³'½"½#½§³
¹ ¶ » ¶
¾ .
°À¿=² If °?³²´pµ¶ , then °À¿=²{·,¸¹'º
¿»8º#³)½#½"½§³¹ ¶ ¿8» ¶¾ .
°«ÁU² If °?³²´pµ¶ , then °0ÁU²·¸
¹+º+Á[»8º-³'½#½"½'³¹ ¶ ÁU» ¶
¾ .
°«ÂU² If °?³²´pµ¶ , then °0ÂU²·¸
¹ º Â[» º ³'½#½"½'³¹ ¶ ÂU» ¶ ¾ .
°'ò In general, if °?³²´µ|¶ , and Ã{ĵ®Å{µÆǵ , then
°'òȷ3¸¹ º Ã
» º ³'½"½"½
³§¹ ¶ Ã
» ¶ ¾ .
É Ãj° É ´µ °´{µ|¶ Ã{Ä8µÊÅ[µÆǵ , then
If É Ã°·3¸ ,É Ã§¹ º ³'½#½",½§and ³ É Ã
¹ ¶ ¾ .
°À˲ If °?³²´pµ¶ , then °À˲{·¹'º
»8ºÍÌʹ§Î"»-Î?̱"±#±'Ìʹ ¶ » ¶ .
°ÅU² If °?³²´pµ¶ , then
°Å«²·3¸¹ Î »"ÏÐʹjÏ#» Î ³¹jÏ-» º Ðȹ º »"ϳ¹ º » Πй Î » º ¾ .
°
Ñ ² If °´{µ|¶ and ²´pµÒ , then
°:Ñ ²·,¸¹'º"³À½"½#½§³¹ ¶ ³»8º³'½"½#½'³» ÒÓ¾ .
а If °´{µ|¶ , then аp·Ô¸:Ð4¹ º ³À½"½"½
³§Ð4¹ ¶ ¾ .
Õ °Ö Õ
If °´{µ|¶ , then °ÖÓ·3¸ ¹'º¤Öj³À½"½#½§³ ¹ ¶ Ö ¾ .
Õ Õ
× °Ø ×
If °´{µ|¶ , then °ØÓ·3¸ ¹ º Øj³À½"½#½§³ ¹ ¶ Ø ¾ .
× ×
Ù °§Ú Ù
If °´{µ|¶ , then °§ÚD·3¸ ¹ º Ú
³'½#½"½§³ ¹ ¶ Ú ¾ .
Ù Ù
ÛjÜ ¸Ý° ¾ If °{·¸
¹ º ³¹ Î ³"½"½#½³:¹ ¶ ¾ ´pµ¶ , then ÛjÜ ¸Ý° ¾ ·1¹ Ü .
Þ° Þ
If °´{µ|¶ , then °·¹'º|Ìʹ§Î?̱"±#±'Ìʹ ¶ .
ß° ß
If °´{µ|¶ , then °[·7¹+º¼¹§Î6±#±"±¹ ¶ .
Á|° If °´{µ|¶ , then Á|°·¹+º+Á[¹§Î?Áà±#±"±
Á¹ ¶ .
Â|° If °´{µ|¶ , then Â|°·¹ º Â[¹ Î Âà±#±"±
¹ ¶ .
á° áÎ á á Î Î
If °´{µ|¶ , then ° Î ·àâ ¹ º Ìã±#±"±
Ìä¹ ¶ .
ᰠẠá á
If °´{µ|¶ , then ° º ·æå ¹'ºåÌ7å ¹§Îå
̱"±"±DÌãå ¹ ¶ å .
á ° á#ç If °´{µ|¶ , then °
á á#ç ·æå ¹+º=åÁå ¹
Î8å
Áà±#±"±
Áæå ¹ ¶ å .
èéÝê ¸° ¾ If °´{µ|¶ , then
èé¤ê ¸¤° ¾ ·ë .
ì Ìí ì
If ³
íïîðµ|¶ , then Ìíñ·3ò#°0Ì®²óÄô°õ´
ì ìö ë è ²´í÷ .
ì Ðí ì
If ³
íïîðµ|¶ , then Ð{íñ·3ò#°UвóÄô°õ´
ì ìö ë è ²´í÷ .
ì Ìø If
ì î1µ|¶ and Û ´{µ|¶ , then ì Ìø·3ò#°0Ì®øÄô°{´ ì ÷ .
© 2001 by CRC Press LLC
Symbol Explanation
ùûúü If
ù]ý1þ|ÿ and þ|ÿ , then ùæú [úü
ü ù
.
ù If
ùïý ðþ|ÿ , then ù ù
.
ù! If
ùïý ðþ|ÿ , then ù!"#$ ù"%'&)(
* .
ù!+, ùïý ðþ|ÿ
ù-+.#$/ , then ù.%'&)(
* ù0 .
If
ù21, If
ùïý ðþ|ÿ , then ù1.#4356798! ù
%:&)(;7 .
úù If
ù]ý1þ|ÿ , then ú"ù 8ú<= ù> .
ù? If
ù]ý1þ|ÿ , then ù" ? # þA@,%'&)(!* ù
.
BC 3Ý-ù 8 ù]ý1þ|ÿ B$C 3Ý-ù 86 supremum of ù . If
ù"ED$F9, G4then
If
HHHIF ÿ 4 then the B$C 3Ý-ù 86JEDLK9G6KH$HHMK, ÿ .
N ù ù order O ,
LP< ù
QR O with
For a N point set
ù
9PSUTVtotal 9P .
W5&)X)ù!3 8 ù]ý1þ|ÿ WY&)XZù-3 86 ù
If
ù" D F , G then H$HHF ÿ , thentheWY&)XZùinfimum
-3 86J ofDL[ .G\If[ H$HH [ ÿ .
] ù ù total order O ,
P ù
QR)^ Owith
For a ] point set
_ UTV ù P .
`a W $` b 3¤!ù 8 ù]ý1þ|ÿ , then
If
`a W `b 3-ù 8 ùc3d%'&)(:efhg `a Bb & b f b e b &ZiI8 .
` %'¤ù(j!3 8 ù]ý1þ|ÿ , then ` %'¤ù(j-3 8Eki aVb` %'('WY&)%'fhWlUùiIgmX . In
If
particular, if
ù
D nnn ÿ , then ` %'¤ù(j!3 8Eo& .
Morphology
Symbol Explanation
£ ¤¥§¦ £ is a function from
¤ into
¦<¨
©'ª«,¬'d® £ The domain of the function
£ ¤¥§¦ is the set ¤
¨
¯ ¬'®±°:² £ £ ¤¥§¦ is the set
£ U³±range
The
:³´!¤µ¢ .
of the function
×dØVÙ
Ñ Øð
×UرÙEÛ
The characteristic function
is defined by
Symbol Explanation
ä <
ä Bold, lowercase characters are used to represent images.
Image variables will usually be chosen from the beginning of
the alphabet.
ð !#
" !
The image is an -valued image on $ . The set
!
is called
the value set of and $ the spatial domain of .
ð !#
% ! Ñ %
Let be a set with unit . Then
whose pixel values are .
Ñ denotes an image, all of
ð !#
& ! &
Let be a set with zero . Then denotes an image, all of
whose pixel values are .
' ( ð"!#
Û
The domain restriction of × ! Ù .! () of
to a subset ð $ is
defined by ' ( +* )-, . Thus, ' ( .
U
/e
Let / t g 4"7 ; /e < @ M t @ M F F N @ M F < /O@ M F e Z t M 4AB. .
¡W¢£ / / g 4@ ¤ F 7 ;
e
¡W¢3£ / t <j
Let
¥ @ M t @ M FF N @ M F < ¡c¢3£ e U Z /O@ Ft M 4AB§¦ .
e
/©¨ If /A4"67 and 6 has a conjugation operation *, then the
pointwise conjugate of image / , /d¨@ M F < @r/O@ M F F .
¨
ª / ª /
ª Nx6denotes reduction by a generic reduce operation
7-
6 (Section 1.4).
The following four items are specific examples of the global reduce operation. Each
assumes /A4"7 and B <K3M w t M y t3|R|3|t M { P .
Symbol Explanation
ÌÎÍ ÏÍCÐ Bold, lowercase characters are used to represent templates.
Usually characters from the middle of the alphabet are used
as template variables.
Image-Template Operations
In the table below, ó is a finite subset of .
Symbol Explanation
ì
í ð ý ì+ý
Let ñ
be a semiring and ñù , ÷Èñsì ùúRû , then
the generic right convolution product of with is defined
ìsívÿ í1 ð ì í1 ð
as írÿððõÿý ò írÿ
ð
ì ìÒý
With the conditions above, except that now
ì ÷ ñ û ú ù , the
generic left convolution product of with is defined as
ì !
írÿ" íWÿð ðbõÿ;ý ò rí ÿ#
ð í%1 ð ì þ ívÿð
$
' & ì * ý ý +.- 0/
Let ò)( , ñù , and ìÒý ÷ ñsù ú û , where ñ , .
The right linear convolution product is defined as
' & ì21
èívÿ3 íWÿð ðbõÿý ò ívÿ465
ð
þ87 í%1 EDð ì í%1 Fð G
ì & ù:98ìÒ;=ý <?>A@CB
With the conditions above, except that ÷Cñ û ú ù , the left
linear convolution product is defined as
ì & )HIJ rí ÿ " íWÿð ðõÿ.ý ò$ íWÿð 5 í%1 EDð ì þ írÿELð M
þ87 í ð N
ù:98; >K@
PQ
O R R
For S
TVUWXY and T[ZUWXY]\^ , the right morphological
max convolution product is defined by
Symbol Explanation
Ð ¨lÑ
Italic uppercase characters are used to denote neighborhoods.
Ñ [Òn
A neighborhood is an image whose pixel values are sets of
¬
points. InÑÔ « ¬ÖÕ×ÒC a neighborhood from to Ó is a
particular,
function .
Ñ,¦?ØÙª ¬
A parameterized neighborhood Ñ) « Ó with parameters
ÒC .
from to
in Ú is a function of the form Ú Õ
Ñ ¸ Ñ ÒC Ñ'¸ ÒC
Ñ'¸¦¶kª
Let §
¢
Û ¬ C
«
, the¶ Ü
Ñ %
¦
transpose
4
§ Þ
ª Ý is defined as
¶ Ñ,¦%§ªßáà'§ ãâ ¸¦%¶kª , that is,
.
Ñ¡äå¾ · Ñdæ Ñ]ä Ñdæ
Ñ,¦§4dilation
The ª ¢ ç of by is defined by
è ¯=éê3ë?ì=íEî3ï]ð$î%ñòó[îAô'õñ4ò3ò .
© 2001 by CRC Press LLC
Image-Neighborhood Operations
Symbol Explanation
ùúÊû ü ú ù ú
þ ùú þ .
If þ
ýÿ , then
"!# ý ùú þ ú , and reduce operation
$% ù&ú
' ù ,, ýÆ
Given þ
defined as (
!# $
,
-
$.0/ ,
1
of with ý is
"!# ý 4
( 56 ý*( )+)7( ) ) þ8 6 ( ) $
( 23. ) . Thus,
the generic right reduction
!# ú
ý
except that now ýÖþ
9
With the conditions above, , the
A( ý
generic !# ) ( !# ý;:<) .
left reduction of with is defined as
>=#
Given
þã' ÷ , and the image average function
ý
? ÷ ÷ , the right reduction of with ý yields the
( @=# ý*)+( ) ?
,0 .
neighborhood averaging operation,
%# A Given
C
ý
B ÷ þã' ÷ ÷ , ,and the image median function
the right reduction of with ý yields the
( # A ý)D( ) B
,- .
neighborhood median filtered image,
Symbol Explanation
E JF , EHF
K The matrix product of matrices E and F .
E># N O: F E F
E># N O: F F&:3# N O EQ: :
The dual p-product of matrices and , defined by
.
References
[1] G. Ritter, “Image algebra.” Unpublished manuscript, available via anonymous ftp from
ftp://ftp.cise.ufl.edu/pub/src/ia/documents, 1994.
1. IMAGE ALGEBRA
1.1. Introduction
1.2. Point Sets
1.3. Value Sets
1.4. Images
1.5. Templates
1.6. Recursive Templates
1.7. Neighborhoods
1.8. The p-Product
1.9. Exercises
1.10. References
4. THRESHOLDING TECHNIQUES
4.1. Introduction
4.2. Global Thresholding
4.3. Semithresholding
4.4. Multilevel Thresholding
4.5. Variable Thresholding
4.6. Threshold Selection Using Mean and Standard Deviation
4.7. Threshold Selection by Maximizing Between-Class Variance
4.8. Threshold Selection Using a Simple Image Statistic
4.9. Exercises
4.10. References