Download presentation
Presentation is loading. Please wait.
1
Chapter 6 Neighborhood Operators
Computer Vision Chapter 6 Neighborhood Operators Presented by: 黃士展 傅楸善 教授 Digital Camera and Computer Vision Laboratory Department of Computer Science and Information Engineering National Taiwan University, Taipei, Taiwan, R.O.C.
2
6.1 Introduction Visit each point p in the image data and do {
N = a neighborhood or region of the image data around the point p result(p) = f(N) } After applying Neighborhood Operation Same dimension Might be interpreted as image Element type based on the functions RAW image DC & CV Lab. CSIE NTU
3
6.1 Introduction Domain type: Numeric, Symbolic.
Neighbor type: 4-neighbor, 8-neighbor, etc. Recursive type: Rec, Non-Rec. DC & CV Lab. CSIE NTU
4
6.1 Introduction: Domain Type
Numeric domain: arithmetic operations, +, -, min, max Symbolic domain: Boolean operations, AND, OR, NOT, table-look-up DC & CV Lab. CSIE NTU
5
6.1 Introduction: Neighbor Type
Neighborhood might be asymmetric DC & CV Lab. CSIE NTU
6
6.1 Introduction: Neighbor Type
(X ± 1, Y) and (X, Y ± 1) (X ± 1 , Y ± 1) (X or X ± 1, Y or Y ± 1) – (X, Y) DC & CV Lab. CSIE NTU
7
6.1 Introduction: Recursive Type
Non-recursive neighborhood operators: output is function of input (Parallel) Recursive neighborhood operators: output depends partly on previous output (Sequential) DC & CV Lab. CSIE NTU
8
6.1 Introduction common masks for noise cleaning
Box filter = unweighted filter DC & CV Lab. CSIE NTU
9
6.1 Introduction common masked for noise cleaning DC & CV Lab.
CSIE NTU
10
6.1 Introduction Different N x N box filter effects comparison
Original Image 3 x 3 5 x 5 9 x 9 15 x 15 35 x 35 DC & CV Lab. CSIE NTU
11
6.1 Introduction 7.69 DC & CV Lab. CSIE NTU
12
6.2 Symbolic Neighborhood Operators
indexing of neighborhoods DC & CV Lab. CSIE NTU
13
6.2.0 Dilation and Erosion (Revisit)
DC & CV Lab. CSIE NTU
14
6.2.0 Dilation DC & CV Lab. CSIE NTU
15
DC & CV Lab. CSIE NTU
16
DC & CV Lab. CSIE NTU
17
DC & CV Lab. CSIE NTU
18
6.2.0 Erosion DC & CV Lab. CSIE NTU
19
6.2.0 Erosion DC & CV Lab. CSIE NTU
20
6.2.0 Erosion DC & CV Lab. CSIE NTU
21
DC & CV Lab. CSIE NTU
22
6.2.0 Opening & Closing DC & CV Lab. CSIE NTU
23
6.2.0 Opening DC & CV Lab. CSIE NTU
24
6.2.0 Opening DC & CV Lab. CSIE NTU
25
6.2.0 Closing DC & CV Lab. CSIE NTU
26
6.2.0 Closing DC & CV Lab. CSIE NTU
27
6.2.0 Application DC & CV Lab. CSIE NTU
28
Opening Closing
29
6.2.1 Region Segmentation Region Segmentation:
Image into many regions Connected, Similarity Grayscale, Color, Texture, etc. Spatial Continuity DC & CV Lab. CSIE NTU
30
6.2.1 Region Segmentation 整張影像的每一像素都必須被分區 每一區塊必須是一個完整的連結區塊
沒有任一像素可以被分類到兩不同區塊 DC & CV Lab. CSIE NTU
31
6.2.1 Region Segmentation 同一區塊內的像素都具備相似的特性 不同特性的像素一定分到不同的區塊
DC & CV Lab. CSIE NTU
32
6.2.1 Region-Growing Operator
: projection, outputs first or second argument : background DC & CV Lab. CSIE NTU
33
6.2.1 Region-Growing Operator
4-connected: output: DC & CV Lab. CSIE NTU
34
6.2.1 Region-Growing Operator
8-connected: output: DC & CV Lab. CSIE NTU
35
6.2.1 Region-Growing Operator
x7 x2 X6 X3 x0 X1 X8 X4 x5 DC & CV Lab. CSIE NTU
36
6.2.1 Region-Growing Operator
Step1. Initial Seed Sets -> New Region Step2. Neighboring pixels similarity. Step3. Pixels remaining -> Step1, or stop DC & CV Lab. CSIE NTU
37
6.2.1 Region-Growing Operator
Seed sets choice is based on user criterion, thus image info is important Grayscale -> Histogram Texture -> Orientation And so on Keep examining another pixels with seed points, classify them. DC & CV Lab. CSIE NTU
38
6.2.1 Region-Growing Operator
Some important issues : The suitable selection of seed points is important More information of the image is better The value, “minimum area threshold” The value, “similarity threshold value“ DC & CV Lab. CSIE NTU
39
6.2.1 Region-Growing Operator
DC & CV Lab. CSIE NTU
40
6.2.1 Region-Growing Operator
Grayscale lightning image. The grayscale value of this image is from 0 to 255 DC & CV Lab. CSIE NTU
41
6.2.1 Region-Growing Operator
All the points with grayscale 255 are considered as seed points DC & CV Lab. CSIE NTU
42
Threshold: (a) >225 (b) >190 (c) >155
DC & CV Lab. CSIE NTU
43
6.2.1 Region-Growing Operator
Advantages: Can provide the original images which have clear edges with good segmentation results. Only need a small number of seed points to grow the region we want. We can determine the seed points and the criteria we want to make. We can choose the multiple criteria at the same time. It performs well with respect to noise. DC & CV Lab. CSIE NTU
44
6.2.1 Region-Growing Operator
The differences between Dilation and Region-Growing are ? Dilation: Initial Area, Structure elements ,Binary Image Region-Growing: Seeds, criterion ,Labeled Regions DC & CV Lab. CSIE NTU
45
Take a Break DC & CV Lab. CSIE NTU
46
6.2.2 Nearest Neighbor Sets and Influence Zones
DC & CV Lab. CSIE NTU
47
6.2.3 Region-Shrinking Operator
Can change all border pixels to background Can change connectivity Can entirely delete region if repeatedly applied DC & CV Lab. CSIE NTU
48
6.2.3 Region-Shrinking Operator
: whether or not arguments identical : background border: has different neighbor and becomes background DC & CV Lab. CSIE NTU
49
6.2.3 Region-Shrinking Operator
4-connected: output: DC & CV Lab. CSIE NTU
50
6.2.3 Region-Shrinking Operator
8-connected: output: DC & CV Lab. CSIE NTU
51
6.2.3 Region-Shrinking Operator
region shrinking: related to binary erosion except on labeled region DC & CV Lab. CSIE NTU
52
6.2.4 Mark-Interior/Border-Pixel Operator
mark-interior/border-pixel operator marks all interior pixels with the label and all border pixels with the label DC & CV Lab. CSIE NTU
53
6.2.4 Mark-Interior/Border-Pixel Operator
: whether or not arguments identical : recognizes whether or not its argument is symbol DC & CV Lab. CSIE NTU
54
6.2.4 Mark-Interior/Border-Pixel Operator
4-connected: output: DC & CV Lab. CSIE NTU
55
6.2.4 Mark-Interior/Border-Pixel Operator
8-connected: output: DC & CV Lab. CSIE NTU
56
6.2.5 Connectivity Number Operator
connectivity number: nonrecursive and symbolic data domain connectivity number: classify the way pixel connected to neighbors six values of connectivity: five for border, one for interior border: isolated, edge, connected, branching, crossing DC & CV Lab. CSIE NTU
57
6.2.5 Connectivity Number Operator
DC & CV Lab. CSIE NTU
58
6.2.5 Connectivity Number Operator
DC & CV Lab. CSIE NTU
59
6.2.5 Connectivity Number Operator
DC & CV Lab. CSIE NTU
60
Yokoi Connectivity Number
4- Connectivity number (= y) DC & CV Lab. CSIE NTU
61
Yokoi Connectivity Number
DC & CV Lab. CSIE NTU
62
Yokoi Connectivity Number
5: all corners are consistent, interior : same as neighbor, but corner DC & CV Lab. CSIE NTU
63
Yokoi Connectivity Number
DC & CV Lab. CSIE NTU
64
Yokoi Connectivity Number
DC & CV Lab. CSIE NTU
65
Yokoi Connectivity Number
lena.64*64 DC & CV Lab. CSIE NTU
66
Yokoi Connectivity Number
DC & CV Lab. CSIE NTU
67
Yokoi Connectivity Number
8-connectivity, only slightly different DC & CV Lab. CSIE NTU
68
Yokoi Connectivity Number
DC & CV Lab. CSIE NTU
69
Rutovitz Connectivity Number
Rutovitz connectivity number: sometimes called crossing number h(a,b,c) = { 1, if (a=b and a != c) or (a != b and a =c) 0, otherwise } a1 = h( X0, X1, X6) a2 = h( X0, X6, X2) a3 = h( X0, X2, X7) a4 = h( X0, X7, X3) a5 = h( X0, X3, X8) a6 = h( X0, X8, X4) a7 = h( X0, X4, X5) a8 = h( X0, X5, X1) DC & CV Lab. CSIE NTU
70
Take a Break DC & CV Lab. CSIE NTU
71
6.2.6 Connected Shrink Operator
connected shrink: recursive operator, symbolic data domain connected shrink: deletes border pixels without disconnecting regions DC & CV Lab. CSIE NTU
72
6.2.6 Connected Shrink Operator
Top-down, left-right scan: deletes edge pixels not right-boundary DC & CV Lab. CSIE NTU
73
DC & CV Lab. CSIE NTU
74
6.2.6 Connected Shrink Operator
: determines whether corner connected DC & CV Lab. CSIE NTU
75
6.2.6 Connected Shrink Operator
: background output symbol DC & CV Lab. CSIE NTU
76
6.2.6 Connected Shrink Operator
DC & CV Lab. CSIE NTU
77
6.2.6 Connected Shrink Operator
h (b, c, d, e) = { 1, if b = c and (d != b or e != b) 0, otherwise } Output symbol y = f (1 , 0 , 0 , 0 , 1) = g DC & CV Lab. CSIE NTU
78
6.2.6 Connected Shrink Operator
g 1 DC & CV Lab. CSIE NTU
79
6.2.7 Skeleton / Medial Axis (Revisit)
DC & CV Lab. CSIE NTU
80
6.2.7 Skeleton / Medial Axis Foreground regions are made of some uniform slow-burning material Light fires simultaneously at all points along the boundary of this region Watch the fire move into the interior Fire meets -> extinguished The so called `quench line‘ which is the skeleton arises. DC & CV Lab. CSIE NTU
81
6.2.7 Skeleton / Medial Axis DC & CV Lab. CSIE NTU
82
6.2.7 Skeleton / Medial Axis DC & CV Lab. CSIE NTU
83
6.2.7 Skeleton / Medial Axis DC & CV Lab. CSIE NTU
84
6.2.7 Skeleton / Medial Axis Take this image for example DC & CV Lab.
CSIE NTU
85
6.2.8 Thinning Operator DC & CV Lab. CSIE NTU
86
6.2.8 Thinning Operator Goal :
Thin the regions in binary image into a one-pixel-width line figure. Thin != Skeleton DC & CV Lab. CSIE NTU
87
6.2.8 Thinning Operator The 4 constraints on thinning operator:
The subtle edge-end must be reserved Connected part couldn’t be broken Connected areas must lead to continuous line Thinned graph must be similar with the medial axis DC & CV Lab. CSIE NTU
88
6.2.8 Thinning Operator Algorithm:
It can be separated into 2 parts A and B. Also, we use 8-neighbor here. DC & CV Lab. CSIE NTU
89
6.2.8 Thinning Operator DC & CV Lab. CSIE NTU
90
6.2.8 Thinning Operator Goal of part A:
Eliminate the up-left corner, right, bottom pixels. DC & CV Lab. CSIE NTU
91
Fig. After applying algo part A to original input image
6.2.8 Thinning Operator Fig. After applying algo part A to original input image DC & CV Lab. CSIE NTU
92
6.2.8 Thinning Operator DC & CV Lab. CSIE NTU
93
6.2.8 Thinning Operator Goal of part B:
Eliminate the bottom-right corner, left, top pixels. DC & CV Lab. CSIE NTU
94
Fig. After applying algo part B to previously output image
6.2.8 Thinning Operator Fig. After applying algo part B to previously output image DC & CV Lab. CSIE NTU
95
Fig. Apply thinning algorithm once again
6.2.8 Thinning Operator Fig. Apply thinning algorithm once again DC & CV Lab. CSIE NTU
96
6.2.8 Thinning Operator During the thinning algo, some rules should be satisfied: Connected regions are thinned to 8-connected paths. Thinning results be as close to skeleton as possible. Extra spurs (short branches) caused by thinning must be minimized. DC & CV Lab. CSIE NTU
97
6.2.8 Thinning Operator DC & CV Lab. CSIE NTU
98
6.2.8 Thinning Operator DC & CV Lab. CSIE NTU
99
Take a Break DC & CV Lab. CSIE NTU
100
6.2.9 Distance Transformation Operator
distance transformation: produces distance to closest border pixel DC & CV Lab. CSIE NTU
101
6.2.9 Distance Transformation Operator
nonrecursive, iterative, start with : background : border, because distance to border is : interior pixels DC & CV Lab. CSIE NTU
102
6.2.9 Distance Transformation Operator
iteration: label with all with neighbor DC & CV Lab. CSIE NTU
103
6.2.9 Distance Transformation Operator
equivalent algorithm: nonrecursive iterative : if no neighbor labeled, still interior, leave it alone : self interior but neighbor labeled : already labeled, won’t change value since later route farther 4-connected: output DC & CV Lab. CSIE NTU
104
6.2.9 Distance Transformation Operator
distance transformation: produces distance to closest background recursive, two-pass DC & CV Lab. CSIE NTU
105
6.2.9 Distance Transformation Operator
first pass: left-right, top-bottom : if background still background, since min: min of upper and left neighbors and add 4-connected output DC & CV Lab. CSIE NTU
106
6.2.9 Distance Transformation Operator
second pass: right-left, bottom-up 4-connected output DC & CV Lab. CSIE NTU
107
6.2.9 Distance Transformation Operator
DC & CV Lab. CSIE NTU
108
6.2.9 Distance Transformation Operator
after pass 1 DC & CV Lab. CSIE NTU
109
6.2.9 Distance Transformation Operator
pass 1 DC & CV Lab. CSIE NTU
110
Take a Break DC & CV Lab. CSIE NTU
111
Radius of Fusion DC & CV Lab. CSIE NTU
112
Radius of Fusion DC & CV Lab. CSIE NTU
113
6.2.11 Number of Shortest Paths
number of shortest paths for each 0-pixel: to binary-1 pixel set given binary image : can be 4-neighborhood or 8-neighborhood : nonzero, stays unchanged since shortest paths counted : if zero then sum of neighboring shortest paths DC & CV Lab. CSIE NTU
114
6.2.11 Number of Shortest Paths
DC & CV Lab. CSIE NTU
115
6.2.11 Number of Shortest Paths
DC & CV Lab. CSIE NTU
116
6.2.11 Number of Shortest Paths
DC & CV Lab. CSIE NTU
117
Take a Break DC & CV Lab. CSIE NTU
118
6.3 Extremum-Related Neighborhood Operators
DC & CV Lab. CSIE NTU
119
6.3.1 Non-Minima-Maxima Operator
non-minima-maxima: nonrecursive operator, symbolic output DC & CV Lab. CSIE NTU
120
6.3.1 Non-Minima-Maxima Operator
a pixel can be neighborhood maximum not relative maximum DC & CV Lab. CSIE NTU
121
6.3.1 Non-Minima-Maxima Operator
DC & CV Lab. CSIE NTU
122
6.3.1 Non-Minima-Maxima Operator
4-connected: output pixel DC & CV Lab. CSIE NTU
123
6.3.2 Relative Extrema Operator
relative extrema operators relative maximum and minimum operators relative extrema recursive operator, numeric data domain input not changed;output modified top-down, left-right scan then bottom-up, right-left scan until no change DC & CV Lab. CSIE NTU
124
6.3.2 Relative Extrema Operator
output value: of highest extrema reachable by monotonic path relative extrema pixels: output same as input pixels DC & CV Lab. CSIE NTU
125
6.3.2 Relative Extrema Operator
pixel designations for the normal and reverse scans DC & CV Lab. CSIE NTU
126
6.3.2 Relative Extrema Operator
DC & CV Lab. CSIE NTU
127
6.3.2 Relative Extrema Operator
two primitive functions and : use new maximum value when ascending : keep original maximum when descending DC & CV Lab. CSIE NTU
128
6.3.2 Relative Extrema Operator
4-connected, output : top-down , left-right : bottom-up , right-left DC & CV Lab. CSIE NTU
129
6.3.2 Relative Extrema Operator
a0 = l0 a1 = h (x0, x1, a0, l1) a2 = h (x0, x2, a1, l2) 3 3,3,3,3 ) x2 , l2 x1 , l1 a0 l0, x0 x2 , l2 x1 , l1 a0 l0, x0 DC & CV Lab. CSIE NTU
130
6.3.2 Relative Extrema Operator
a0 = l0 a1 = h (x0, x1, a0, l1) a2 = h (x0, x2, a1, l2) 5 5, 5, 5, 5) x2 , l2 x1 , l1 a0 l0, x0 x2 , l2 x1 , l1 a0 l0, x0 DC & CV Lab. CSIE NTU
131
6.3.2 Relative Extrema Operator
DC & CV Lab. CSIE NTU
132
6.3.2 Relative Extrema Operator
DC & CV Lab. CSIE NTU
133
6.3.2 Relative Extrema Operator
DC & CV Lab. CSIE NTU
134
6.3.2 Relative Extrema Operator
DC & CV Lab. CSIE NTU
135
6.3.2 Relative Extrema Operator
4 DC & CV Lab. CSIE NTU
136
6.3.2 Relative Extrema Operator
a0 = l0 a1 = h (x0, x1, a0, l1) a2 = h (x0, x2, a1, l2) 6 x2 1 9 1 9 x1 l1 a0 l0 x0 l2 x0 2 x2 x1 l2 a0 l0 DC & CV Lab. CSIE NTU
137
6.3.2 Relative Extrema Operator
DC & CV Lab. CSIE NTU
138
6.3.3 Reachability Operator
successively propagating labels that can reach by monotonic paths descending reachability operator employs DC & CV Lab. CSIE NTU
139
6.3.3 Reachability Operator
g: pixels that are not relative extrema labeled background a: unchanged when neighbor is g or the same with neighbor b: become first propagated label if originally g c: common region more than one extremum can reach it by monotonic path c: already labeled and neighbor has different label, then two extrema reach a: propagate from extremum DC & CV Lab. CSIE NTU
140
6.3.3 Reachability Operator
4-connected, output a0 = l0 a1 = h (a0 , l1 , x0 , x1) a2 = h (a1 , l2 , x0 , x2) DC & CV Lab. CSIE NTU
141
6.3.3 Reachability Operator
DC & CV Lab. CSIE NTU
142
6.3.3 Reachability Operator
DC & CV Lab. CSIE NTU
143
6.3.3 Reachability Operator
DC & CV Lab. CSIE NTU
144
Take a Break DC & CV Lab. CSIE NTU
145
6.4 Linear Shift-Invariant Neighborhood Operators
convolution: commutative, associative, distributor over sums, homogeneous DC & CV Lab. CSIE NTU
146
6.4.1 Convolution and Correlation
convolution of an image f with kernel w DC & CV Lab. CSIE NTU
147
6.4.1 Convolution and Correlation
DC & CV Lab. CSIE NTU
148
6.4.1 Convolution and Correlation
DC & CV Lab. CSIE NTU
149
6.4.1 Convolution and Correlation
DC & CV Lab. CSIE NTU
150
6.4.1 Convolution and Correlation
DC & CV Lab. CSIE NTU
151
6.4.1 Convolution and Correlation
DC & CV Lab. CSIE NTU
152
6.4.1 Convolution and Correlation
DC & CV Lab. CSIE NTU
153
6.4.1 Convolution and Correlation
DC & CV Lab. CSIE NTU
154
6.4.1 Convolution and Correlation
DC & CV Lab. CSIE NTU
155
6.4.2 Separability straightforward computation:
multiplications, additions decomposition: DC & CV Lab. CSIE NTU
156
6.4.2 Separability DC & CV Lab. CSIE NTU
157
Project due Nov. 6 Write a program to generate Yokoi connectivity number You can down sample lena.bmp from 512*512 to 64*64 first. Sample pixels positions at (r, c) = (0,0), (0,8) ……, so everyone will get the same answer . DC & CV Lab. CSIE NTU
158
Project due Nov. 20 Write a program to generate thinned image.
You can down sample lena.bmp from 512*512 to 64*64 first. Sample pixels positions at (r, c) = (0,0), (0,8) ……, so everyone will get the same answer . DC & CV Lab. CSIE NTU
159
Midterm Nov. 13 extent: whatever covered up to Nov. 6 (inclusive)
DC & CV Lab. CSIE NTU
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.