Download presentation
Presentation is loading. Please wait.
Published byJonas Clifton Berry Modified over 9 years ago
1
MATHEMATICS OF BINARY MORPHOLOGY APPLICATIONS IN Vision
and APPLICATIONS IN Vision
3
In general; what is “Morphology”?
The science of form and structure the science of form, that of the outer form, inner structure, and development of living organisms and their parts about changing/counting regions/shapes Among other applications it is used to pre- or post-process images via filtering, thinning and pruning Smooth region edges create line drawing of face Force shapes onto region edges curve into a square Count regions (granules) number of black regions Estimate size of regions area calculations
4
What is Morphology in computer vision ?
Morphology generally concerned with shape and properties of objects. Used for segmentation and feature extraction. Segmentation = used for cleaning binary objects. Two basic operations erosion (opening) dilation (closing)
5
Morphological operations and algebras
Different definitions in the textbooks Different implementations in the image processing programs. The original definition, based on set theory, is made by J. Serra in 1982. Defined for binary images - binary operations (boolean, set-theoretical) Can be used on grayscale images - multiple-valued logic operations
6
Morphological operations on a PC
Various but slightly different implementations in Scion Paint Shop Pro Adope Photoshop Corel Photopaint mm Try them, it is a lot of fun
7
Mathematical Morphology - Set-theoretic representation for binary shapes
8
Binary Morphology Morphological operators are used to prepare binary (thresholded) images for object segmentation/recognition Binary images often suffer from noise (specifically salt-and-pepper noise) Binary regions also suffer from noise (isolated black pixels in a white region). Can also have cracks, picket fence occlusions, etc. Dilation and erosion are two binary morphological operations that can assist with these problems.
9
Goals of morphological operations:
1. Simplifies image data 2. Preserves essential shape characteristics 3. Eliminates noise 4. Permits the underlying shape to be identified and optimally reconstructed from their distorted, noisy forms
10
What is the mathematical morphology ?
1. An approach for processing digital image based on its shape 2. A mathematical tool for investigating geometric structure in image The language of morphology is set theory. Mathematical morphology is extension to set theory.
11
Importance of Shape in Processing and Analysis
Shape is a prime carrier of information in machine vision For instance, the following directly correlate with shape: identification of objects object features assembly defects
12
Binary Morphology
13
Shape Operators Set difference Set intersection
Shapes are usually combined by means of : Set Union (overlapping objects): Set Difference based on Set intersection (occluded objects): Set difference Set intersection
14
Morphological Operations based on combining base operations
The primary morphological operations are dilation and erosion More complicated morphological operators can be designed by means of combining erosions and dilations We will use combinations of union, complement, intersection, erosion, dilation, translation... Let us illustrate them and explain how to combine
15
Libraries of Structuring Elements
Application specific structuring elements created by the user
16
Notation B X x -2 -1 0 1 2 -2 -1 0 1 2 y A special set :
B y A special set : the structuring element Origin at center in this case, but not necessarily centered nor symmetric X No necessarily compact nor filled 3*3 structuring element, see next slide how it works
17
Dilation
18
Explanation of Dilation
Dilation : x = (x1,x2) such that if we center B on them, then the so translated B intersects X. X difference dilation B
19
Notation for Dilation Dilation : x = (x1,x2) such that if we center B on them, then the so translated B intersects X. How to formulate this definition ? 1) Literal translation Mathematical definition of dilation 2) Better : from Minkowski’s sum of sets Another Mathematical definition of dilation uses the concept of Minkowski’s sum B is ingeneral not the same as B
20
The Concept of Minkowski Sum
21
Minkowski’s Sum Definition of Minkowski’s sum of sets S and B : l l
22
Another View at Dilation
23
Comparison of Dilation and Minkowski sum
Bx = x and b are points Minkowski sum
24
It is like dilation but we are not going around , we go only to top and to right
25
Dilation and Minkowski Set
Dilation and Minkowski Set are denoted by + or by No unified notation
26
Dilation is not the Minkowski’s sum
27
Dilation is not the Minkowski’s sum
b b b b l Dilation l B is not the same as B
28
Dilation with other structuring elements
29
Dilation with other structuring elements
30
Dilation vs SE Erosion shrinks Dilation expands binary regions
Can be used to fill in gaps or cracks in binary images structuring Element ( SE ) If the point at the origin of the structuring element is set in the underlying image, then all the points that are set in the SE are also set in the image Basically, its like OR’ing the SE into the image
31
Dilation fills holes Fills in holes. Smoothes object boundaries.
Adds an extra outer ring of pixels onto object boundary, ie, object becomes slightly larger.
32
Main Applications of Dilation
33
Dilation example
34
Possible problems with Morphological Operators
Erosion and dilation clean image but leave objects either smaller or larger than their original size. Opening and closing perform same functions as erosion and dilation but object size remains the same.
35
More Erode and Dilate Examples
Input Image Dilated Eroded Made in Paint Shop Pro
36
Dilation explained pixed by pixel
Denotes origin of B i.e. its (0,0) Denotes origin of A i.e. its (0,0) B • • • A
37
Dilation explained by shape of A
Shape of A repeated without shift B • • • Shape of A repeated with shift A
38
Properties of Dilation
objects are light (white in binary) Dilation does the following: 1. fills in valleys between spiky regions 2. increases geometrical area of object 3. sets background pixels adjacent to object's contour to object's value 4. smoothes small negative grey level regions
39
Structuring Element for Dilation
Length 6 Length 5
40
Structuring Element for Dilation
41
Structuring Element for Dilation
Single point in Image replaced with this in the Result
42
Structuring Element for Dilation
43
Definition of Dilation: Mathematically
Dilation is the operation that combines two sets using vector addition of set elements. Let A and B are subsets in 2-D space. A: image undergoing analysis, B: Structuring element, denotes dilation
44
Dilation versus translation
Let A be a Subset of and The translation of A by x is defined as: The dilation of A by B can be computed as the union of translation of A by the elements of B x is a vector
45
Dilation versus translation, illustrated
Element (0,0) Shift vector (0,1) Shift vector (0,0) • • • B •
46
Dilation using Union Formula
Center of the circle This circle will create one point This circle will create no point
47
Example of Dilation with various sizes of structuring elements
Pablo Picasso, Pass with the Cape, 1960
48
Mathematical Properties of Dilation
Commutative Associative Extensivity Dilation is increasing Illustrated in next slide
49
Illustration of Extensitivity of Dilation
• • B • Replaced with Here 0 does not belong to B and A is not included in A B
50
More Properties of Dilation
Translation Invariance Linearity Containment Decomposition of structuring element
51
Dilation The dilation operator takes two pieces of data as input
A binary image, which is to be dilated A structuring element (or kernel), which determines the behavior of the morphological operation Suppose that X is the set of Euclidean coordinates of the input image, and K is the set of coordinates of the structuring element Let Kx denote the translation of K so that its origin is at x. The DILATION of X by K is simply the set of all points x such that the intersection of Kx with X is non-empty
52
Dilation Example: Suppose that the structuring element is a 3x3 square with the origin at its center 1 { (-1,-1), (0,-1), (1,-1), (-1,0), (0,0), (1,0), ( 1,1), (0,1), (1,1) } K = X =
53
Dilation Example: Suppose that the structuring element is a 3x3 square with the origin at its center Note: Foreground pixels are represented by a color and background pixels are empty
54
Dilation Structuring element Input output
55
Dilation output
56
Dilation output
57
Dilation
58
Dilation
59
Erosion
60
Example of Erosion Erosion : x = (x1,x2) such that if we center B on them, then the so translated B is contained in X. difference erosion
61
Minkowski’s substraction
Notation for Erosion Erosion : x = (x1,x2) such that if we center B on them, then the so translated B is contained in X. How to formulate this definition ? 1) Literal translation Erosion 2) Better : from Minkowski’s substraction of sets Minkowski’s substraction
62
Notation for Erosion Minkowski’s substraction of sets Erosion
BINARY MORPHOLOGY Minkowski’s substraction of sets Erosion
63
Erosion with other structuring elements
64
Erosion with other structuring elements
Did not belong to X When the new SE is included in old SE then a larger area is created
65
Common structuring elements shapes
= origin x y disk circle segments 1 pixel wide Note that here : points
66
Problem in BINARY MORPHOLOGY
using Minkowski Sum First we calculate the operation in parentheses to obtain a diamond
67
PROBLEM BINARY MORPHOLOGY
next we calculate the external operation to obtain a hexagon
68
ANOTHER EXAMPLE OF EROSION
Where d is a diameter Problem : <d/2 d/2 d
69
Implementation of dilation: very low computational cost
1 (or >0) Logical or
70
Implementation of erosion: very low computational cost
1 Logical and
71
More on Erosion
72
Erosion (Minkowski subtraction)
The contraction of a binary region (aka, region shrinking) Use a structuring element on binary image data to produce a new binary image Structuring elements (SE) are simply patterns that are matched in the image It is useful to explain operation of erosion and dilation in different ways.
73
Typical Uses of Erosion
Removes isolated noisy pixels. Smoothes object boundary. Removes the outer layer of object pixels, ie, object becomes slightly smaller.
74
Properties of Erosion:
Erosion removes spiky edges objects are light (white in binary) decreases geometrical area of object sets contour pixels of object to background value smoothes small positive grey level regions
75
Erosion Example
76
Erosion explained pixel by pixel
• • • B
77
Structuring Element in Erosion Example
78
How It Works? During erosion, a pixel is turned on at the image pixel under the structuring element origin only when the pixels of the structuring element match the pixels in the image Both ON and OFF pixels should match. This example erodes regions horizontally from the right.
79
Structuring Element in Erosion Example
80
Structuring Element in Erosion Example
81
Structuring Element in Erosion Example
82
Structuring Element in Erosion Example
83
Structuring Element in Erosion Example
84
Structuring Element in Erosion Example
85
Mathematical Definition of Erosion
Erosion is the morphological dual to dilation. It combines two sets using the vector subtraction of set elements. Let denotes the erosion of A by B
86
Erosion in terms of other operations:
Erosion can also be defined in terms of translation In terms of intersection Observe that vector here is negative
87
Reminder - this was A •
88
Erosion illustrated in terms of intersection and negative translation
Observe negative translation Because of negative shift the origin is here • • • •
89
Erosion formula and intuitive example
Center of B is here and adds a point Center here will not add a point to the Result
90
Example of Erosions with various sizes of structuring elements
Pablo Picasso, Pass with the Cape, 1960
91
Properties of Erosion Erosion is not commutative! Extensivity Erosion is dereasing Chain rule The chain rule is as sharp operator in Cube Calculus used in logic synthesis. There are more similarities of these algebras
92
Properties of Erosion Translation Invariance Linearity Containment
Decomposition of structuring element
93
Duality Relationship between erosion and dilation
Dilation and Erosion transformation bear a marked similarity, in that what one does to image foreground and the other does for the image background. , the reflection of B, is defined as Erosion and Dilation Duality Theorem Observe negative value which is mirror image reflection of B Similar but not identical to De Morgan rule in Boolean Algebra
94
Erosion as Dual of Dilation
Erosion is the dual of dilation i.e. eroding foreground pixels is equivalent to dilating the background pixels.
95
Duality Relationship between erosion and dilation
Easily visualized on binary image Template created with known origin Template stepped over entire image similar to correlation Dilation if origin == 1 -> template unioned resultant image is large than original Erosion only if whole template matches image origin = 1, result is smaller than original Another look at duality
96
One more view at Erosion with examples
To compute the erosion of a binary input image by the structuring element For each foreground pixel superimpose the structuring element If for every pixel in the structuring element, the corresponding pixel in the image underneath is a foreground pixel, then the input pixel is left as it is Otherwise, if any of the corresponding pixels in the image are background, however, the input pixel is set to background value
97
Erosion
98
Erosion example with dilation and negation
We want to calculate this We dilate with negation
99
Erosion .. And we negate the result
We obtain the same thing as from definition
100
Morphological Operations in terms of more general neighborhoods
This exists in Matlab
101
Erode and Dilate in terms of more general neighborhoods
Yet another loook at Duality Relationship between erosion and dilation
102
Edge detection or Binary Contour
103
Boundary Extraction
104
Erode and Binary Contour in Matlab
Erosion can be used to find contour Dilation can be also used for it - think how?
105
Edge detection This subtraction is set theoretical Dilate - original
Now you need to invert the image There are more methods for edge detection
106
Opening & Closing Opening and Closing are two important operators from mathematical morphology They are both derived from the fundamental operations of erosion and dilation They are normally applied to binary images
107
Open and Close Close = Dilate next Erode Open = Erode next Dilate
Original image eroded dilated dilated eroded Open Close
108
OPENING
109
OPENING OPENING Opening : also Supresses : small islands
ithsmus (narrow unions) narrow caps difference
110
Opening with other structuring elements
111
Open An erosion followed by a dilation It serves to eliminate noise
Does not significantly change an object’s size
112
Comparison of Opening and Erosion
Opening is defined as an erosion followed by a dilation using the same structuring element The basic effect of an opening is similar to erosion Tends to remove some of the foreground pixels from the edges of regions of foreground pixels Less destructive than erosion The exact operation is determined by a structuring element.
113
Opening Example Original
What combination of erosion and dilation gives: cleaned binary image object is the same size as in original Original
114
Opening Example Cont Original Erode Dilate Erode original image.
Dilate eroded image. Smooths object boundaries, eliminates noise (isolated pixels) and maintains object size. Original Erode Dilate
115
One more example of Opening
Erosion can be used to eliminate small clumps of undesirable foreground pixels, e.g. “salt noise” However, it affects all regions of foreground pixels indiscriminately Opening gets around this by performing both an erosion and a dilation on the image
116
CLOSING
117
EXAMPLE OF CLOSING Closing : also Supresses : small lakes (holes)
channels (narrow separations) narrow bays
118
Closing previous image with other structuring elements
BINARY MORPHOLOGY With bigger rectangle like this With smaller cross like this
119
Application : shape smoothing and noise filtering
Papilary lines recognition
120
segmentation of microstructures (Matlab Help)
Application : segmentation of microstructures (Matlab Help) opening and with threshold disk radius 6 closing threshold original negated
121
PROPERTIES IN BINARY MORPHOLOGY
all of them are increasing : opening and closing are idempotent :
122
EXTENSIVE VERSUS ANTI-EXTENSIVE OPERATIONS
dilation and closing are extensive operations erosion and opening are anti-extensive operations:
123
DUALITIES OF MORPHOLOGICAL OPERATORS
duality of erosion-dilation, opening-closing,...
124
Decomposition of structuring elements
operations with big structuring elements can be done by a succession of operations with small s.e’s
125
HIT-OR-MISS Hit-or-miss : Bi-phase structuring element “Hit” part
(white) “Miss” part (black)
126
HIT or MISS FOR ISOLATED POINTS
Looks for pixel configurations : background foreground doesn’t matter
127
ISOLATED POINTS isolated points at 4 connectivity
128
More examples on Closing
129
Close Dilation followed by erosion
Serves to close up cracks in objects and holes due to pepper noise Does not significantly change object size
130
More examples of Closing
What combination of erosion and dilation gives: cleaned binary image object is the same size as in original Original
131
More examples of Closing cont
Dilate original image. Erode dilated image. Smooths object boundaries, eliminates noise (holes) and maintains object size. Original Dilate Erode
132
Closing as dual to Opening
Closing, like its dual operator opening, is derived from the fundamental operations of erosion and dilation. Normally applied to binary images Tends to enlarge the boundaries of foreground regions Less destructive of the original boundary shape The exact operation is determined by a structuring element.
133
Closing is opening in revers
Closing is opening performed in reverse. It is defined simply as a dilation followed by an erosion using the same
134
One more example of Closing
135
Mathematical Definitions of Opening and Closing
Opening and closing are iteratively applied dilation and erosion Opening Closing
136
Relation of Opening and Closing
Difference is only in corners
137
Opening and Closing are idempotent
Their reapplication has not further effects to the previously transformed result
138
Properties of Opening and Closing
Translation invariance Antiextensivity of opening Extensivity of closing Duality
139
Example of Openings with various sizes of structuring elements
Pablo Picasso, Pass with the Cape, 1960
140
Example of Closings with various sizes of structuring elements
141
Thinning and Thickening
142
Thinning and Thickening
Thickenning : Depending on the structuring elements (actually, series of them), very different results can be achieved : Prunning Skeletons Zone of influence Convex hull ...
143
Prunning at 4 connectivity
Prunning at 4 connectivity : remove end points by a sequence of thinnings This point is removed with dark green neighbors 1 iteration =
144
IDEMPOTENCE shown as a result of thinning
1st iteration 2nd iteration 3rd iteration: idempotence
145
Other thinning operations
background doesn’t matter foreground
146
USING EROSION TO FIND CONTOURS
Contours of binary regions : erosion difference Contour found with larger mask
147
CONTOURS with different connectivity patterns
Important for perimeter computation.
148
Use of thickening: Convex hull
ii. Convex hull : union of thickenings, each up to idempotence Original shaper Thickening with first mask Union of four thickenings
149
Example of using convex hull
150
iii. Skeleton : Maximal disk : disk centered at x, Dx, such that
Dx X and no other Dy contains it . Skeleton : union of centers of maximal disks.
151
PROBLEMS with skeletons
Instability : infinitessimal variations in the border of X cause large deviations of the skeleton not necessarily connex even though X connex good approximations provided by thinning with special series of structuring elements
152
Example of iterative thinning with 8 masks
1st iteration
153
Example of iterative thinning with 8 masks
result of 1st iteration 2nd iteration reaches idempotence
154
Thinning with thickening 20 iterations of thinning color white
Some sort of region clustering
155
Skeletonization for OCR
BINARY MORPHOLOGY Application : skeletonization for OCR by graph matching skeletonization vectorization
156
skeletonization Application : skeletonization for OCR by graph matching and 3 rotations Hit-or-Miss
157
Calculation of Geodesic zones of influence (GZI)
X set of n connex components {Xi}, i=1..n . The zone of influence of Xi , Z(Xi) , is the set of points closer to some point of Xi than to a point of any other component. Also, Voronoi partition. Dual to skeleton.
158
Calculating and using Geodesic Zones of Influence
thr erosion 7x7 and GZI opening 5x5
159
Calculating and using Geodesic Zones of Influence (cont)
thr erosion 7x7 dist
160
Skeleton by Maximal Balls
161
Example – Morphological Processing of Handwritten Digits
thresholding thinning smoothing opening
162
OF MORPHOLOGICAL OPERATIONS
PROGRAMMING OF MORPHOLOGICAL OPERATIONS
163
USING LISP
164
USING LISP
165
USING LISP
166
USING LISP
167
USING LISP
168
Morphological Filtering
169
Morphological Filtering
Main idea of Morphological Filtering: Examine the geometrical structure of an image by matching it with small patterns called structuring elements at various locations By varying the size and shape of the matching patterns, we can extract useful information about the shape of the different parts of the image and their interrelations. Combine set-theoretical and morphological operations:
170
Example 1: Morphological filtering
Noisy image will break down OCR systems Noisy image Clean original image
171
Morphological filtering (MF)
By applying MF, we increase the OCR accuracy! Restored image
172
Rank Filter Median Input 1 operation 2 operations
173
Postprocessing Opening followed by closing.
Removes noise and smoothes boundaries.
174
Postprocessing Opening followed by closing.
Removes noise and smoothes boundaries.
176
Grey Level Morphology
184
erosion dilation
187
Change of histogram as a result of dilation
195
Removal of Border Objects
Marker is the border itself
196
Summary on Morphological Approaches
Mathematical morphology is an approach for processing digital image based on its shape The language of morphology is set theory The basic morphological operations are erosion and dilation Morphological filtering can be developed to extract useful shape information Methods can be extended to more values and more dimensions Nice mathematics can be formulated - non-linear
197
Conclusion Segmentation separates an image into regions.
Use of histograms for brightness based segmentation. Peak corresponds to object. Height of peak corresponds to size of object. If global image histogram is multimodal, local image region histogram may be bimodal. Local thresholds can give better segmentation.
198
Conclusion Postprocessing uses morphological operators.
Same as convolution only use Boolean operators instead of multiply and add. Erosion clears noise, makes smaller. Dilation fills in holes, makes larger. Postprocessing Opening and closing to clean binary images. Repeated erosion with special rule produces skeleton.
199
Problems 1 - 6 1. Write LISP or C++ program for dilation of binary images 2. Modify it to do erosions (few types) 3. Modify it to perform shift and exor operation and shift and min operation 4. Generalize to multi-valued algebra 5. Create a comprehensive theory of multi-valued morphological algebra and its algorithms (publishable). 6. Write a program for inspection of Printed Circuit Boards using morphological algebra.
200
Problem 7. Electric Outlet Extraction has been done using a combination of Canny Edge Detection and Hough Transforms Write a LISP program that will use only basic morphological methods for this application.
201
Image Processing for electric outlet, how?
Currently there are many, many ways to approach this problem Segmentation Edge Detection DPC compression FFT IFFT DFT Thinning Growing Haar Transform Hex Rotate Alpha filtering DPC compression Perimeter Fractal Gaussian Filter Band Pass Filter Homomorphic Filtering Contrast Sharper Least Square Restoration Warping Dilation
202
Create morphological equivalents of other image processing methods.
Image Processing, how? Create morphological equivalents of other image processing methods. New, publishable, use outlet problem as example to illustrate
203
Problem 8. Openings and Closings as examples.
The solution here is to follow up one operation with the other. An opening is defined as an erosion operation followed by dilation using the same structuring element. Similarly, a closing is dilation followed by erosion. Define and implement other combined operations.
204
Problems 9. Generalize binary morphological algebra from 2 dimensional to 3 dimensional images. What are the applications. 10. Write software for 9. 11. Generalize your generalized multi-valued morphological algebra to 3 or more dimensions, theoretically, find properties and theorems like those from this lecture. 12. Write software for 11.
205
Problem 13 Mathematical morphology uses the concept of structuring elements to analyze image features. A structuring element is a set of pixels in some arrangement that can extract shape information from an image. Typical structuring elements include rectangles, lines, and circles. Think about other structuring elements and their applications.
207
Morphological Operations: Matlab
BWMORPH Perform morphological operations on binary image. BW2 = BWMORPH(BW1,OPERATION) applies a specific morphological operation to the binary image BW1. BW2 = BWMORPH(BW1,OPERATION,N) applies the operation N times. N can be Inf, in which case the operation is repeated until the image no longer changes. OPERATION is a string that can have one of these values: 'close' Perform binary closure (dilation followed by erosion) 'dilate' Perform dilation using the structuring elementones(3) 'erode' Perform erosion using the structuring elementones(3) 'fill' Fill isolated interior pixels (0's surrounded by1's) 'open' Perform binary opening (erosion followed bydilation) 'skel' With N = Inf, remove pixels on the boundariesof objects without allowing objects to break apart demos/demo9morph/
208
Sources Shreekanth Mandayam D.A. Forsyth, University of New Mexico,
Qigong Zheng, Language and Media Processing Lab Center for Automation Research University of Maryland College Park October 31, 2000 John Miller Matt Roach J. W. V. Miller and K. D. Whitehead The University of Michigan-Dearborn Spencer Lustor Light Works Inc. C. Rössl, L. Kobbelt, H.-P. Seidel, Max-Planck Institute for, Computer Science, Saarbrücken, Germany LBA-PC4 Howard Schultz Shreekanth Mandayam ECE Department Rowan University D.A. Forsyth, University of New Mexico
209
More recent Sources Howie Choset G.D. Hager, Z. Dodds, Dinesh Mocha
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.