Introduction to Image Processing
Earl F. Glynn
Scientific Programmer
Bioinformatics
8 March 2002
1
Introduction to Image Processing
• Pixels: Grey Scale
•
•
•
•
Pixels: Color
Color Issues
Image Manipulation
Image Enhancement
2
1
Introduction to Image Processing
Pixels: Grey Scale
1.0 (white)
255 – 8 bits
16383 – 14 bits
65535 – 16 bits
0 (black)
24 Columns
106
107
112
110
115
118
118
117
113
122
119
105
113
117
104
76
115
128
89
98
110
118
120
118
130
120
122
125
115
119
115
122
113
105
117
129
87
95
115
118
123
124
124
129
123
129
126
116
123
124
116
111
117
118
108
107
110
118
123
121
127
130
126
133
109
119
132
118
117
116
113
116
122
118
121
118
120
123
124
116
115
117
96
109
122
123
121
124
112
112
118
121
129
118
121
117
103
108
110
108
107
101
125
121
120
131
117
108
123
129
124
120
119
103
91
86
100
101
101
100
117
107
117
128
123
117
131
138
123
116
106
102
68
53
68
65
66
78
96
108
119
127
128
118
137
133
117
106
101
82
46
46
44
42
28
36
85
104
118
131
125
121
136
128
111
90
78
47
38
30
45
39
20
25
55
103
109
120
135
130
132
127
103
105
58
42
38
40
51
43
28
24
49
92
112
126
130
132
135
119
96
106
63
36
75
185
175
95
56
34
52
82
115
124
123
126
135
118
93
103
42
37
137
213
210
141
117
35
43
100
115
130
127
128
136
122
96
117
63
41
98
194
195
98
40
42
47
99
116
130
129
125
136
123
97
106
62
39
32
84
90
32
35
37
64
90
108
126
118
116
129 132
126 126
101 111
88 106
58 75
52 76
37 51
121 68
199 129
59 44
38 46
46 89
86 95
95 119
116 123
129 136
130 134
129 126
126
134
123
100
87
86
89
87
84
66
86
85
113
121
126
131
133
130
110
128
120
119
99
94
96
88
99
92
86
108
121
125
126
125
127
128
104
117
114
120
116
112
106
116
123
96
96
117
127
129
121
125
126
123
89
99
118
123
130
120
116
116
123
116
115
126
131
120
127
122
125
122
18 Rows
Matrix: 24 Columns x 18 Rows = 432 numbers
86
86
109
122
125
121
120
114
127
124
124
127
127
117
125
117
123
116
82
94
116
116
122
121
117
121
129
121
126
129
125
117
125
121
117
121
99
101
111
102
126
125
115
119
114
115
118
119
122
112
118
118
123
117
Each number
is a picture element
or “pixel”
3
Introduction to Image Processing
Pixels: Grey Scale
0 (black)
24 Columns
1.0 (white)
255 – 8 bits
16383 – 14 bits
65535 – 16 bits
Each square is
a picture element,
or “pixel”
18 Rows
24 Columns x 18 Rows = 432 pixels
Not all display levels are perceivable on all devices
4
2
Introduction to Image Processing
Pixels: Grey Scale
106
107
112
110
115
118
118
117
113
122
119
105
113
117
104
76
115
128
89
98
110
118
120
118
130
120
122
125
115
119
115
122
113
105
117
129
87
95
115
118
123
124
124
129
123
129
126
116
123
124
116
111
117
118
108
107
110
118
123
121
127
130
126
133
109
119
132
118
117
116
113
116
122
118
121
118
120
123
124
116
115
117
96
109
122
123
121
124
112
112
118
121
129
118
121
117
103
108
110
108
107
101
125
121
120
131
117
108
123
129
124
120
119
103
91
86
100
101
101
100
117
107
117
128
123
117
131
138
123
116
106
102
68
53
68
65
66
78
96
108
119
127
128
118
137
133
117
106
101
82
46
46
44
42
28
36
85
104
118
131
125
121
136
128
111
90
78
47
38
30
45
39
20
25
55
103
109
120
135
130
132
127
103
105
58
42
38
40
51
43
28
24
49
92
112
126
130
132
135
119
96
106
63
36
75
185
175
95
56
34
52
82
115
124
123
126
135
118
93
103
42
37
137
213
210
141
117
35
43
100
115
130
127
128
136
122
96
117
63
41
98
194
195
98
40
42
47
99
116
130
129
125
136
123
97
106
62
39
32
84
90
32
35
37
64
90
108
126
118
116
129
126
101
88
58
52
37
121
199
59
38
46
86
95
116
129
130
129
132
126
111
106
75
76
51
68
129
44
46
89
95
119
123
136
134
126
126
134
123
100
87
86
89
87
84
66
86
85
113
121
126
131
133
130
110
128
120
119
99
94
96
88
99
92
86
108
121
125
126
125
127
128
104
117
114
120
116
112
106
116
123
96
96
117
127
129
121
125
126
123
89
99
118
123
130
120
116
116
123
116
115
126
131
120
127
122
125
122
86
86
109
122
125
121
120
114
127
124
124
127
127
117
125
117
123
116
82
94
116
116
122
121
117
121
129
121
126
129
125
117
125
121
117
121
99
101
111
102
126
125
115
119
114
115
118
119
122
112
118
118
123
117
Contour / surface map created using MathCad
5
Introduction to Image Processing
• Pixels: Grey Scale
• Pixels: Color
• Color Issues
• Image Manipulation
• Image Enhancement
6
3
Introduction to Image Processing
Pixels: Color
165
178
191
177
186
192
192
192
194
184
211
206
177
190
188
195
145
195
225
166
153
173
184
204
202
198
208
204
206
205
206
194
188
213
192
189
207
215
167
136
167
190
204
210
206
211
213
207
219
214
201
209
208
207
188
192
213
168
186
176
186
201
209
210
209
220
216
223
200
207
225
207
206
196
191
200
169
202
208
200
207
211
206
210
201
198
202
182
187
212
223
218
216
201
199
170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188
203 221 228 229 222 218 230 229 220 215 209 207 203 169 156 141 116 127 159
207 221 239 231 216 215 201 211 212 211 206 209 217 205 184 145 140 152 161
214 222 215 219 188 176 168 157 164 159 150 186 204 200 180 197 185 188 195
204 207 202 188 152 172 178 169 191 170 134 163 166 183 203 213 204 203 172
207 201 189 164 142 98 116 68 102 106 84 121 148 163 202 218 216 207 212
203 189 171 142 79 68 45 47 49 44 61 105 129 146 175 201 201 202 210
186 152 129 57 52 53 63 143 121 39 40 71 145 144 181 195 199 199 186
177
165155
166 80
167 62
168 49
169 47
170187
171209
172215
173108
174109
175 60
176131
177120
178181
179195
180196
181197
182191
183 184
194
13215796106
108 58
127 55
145 63
131196
139212
153217
160106
159211
159124
157107
165118
162175
170194
162202
157210
148186
129 128
194
115168
105 96
107 57
128 51
129 57
141113
147169
153132
153 45
136 69
142 44
136107
130133
139160
138200
147206
146204
165193
152 138
175
136178
122113
131 43
124 39
135 45
150 50
137141
139 50
124 50
121 47
109 59
106125
115146
109158
116192
110210
120210
151202
130 131
178
131160
139140
136 58
140 41
137 31
135 44
146 44
136 54
118 49
106 59
125134
131125
131158
151189
130204
104207
121209
111194
143 137
217
196
168
141
86
64
72
58
65
90
128
152
188
199
217
217
220
210
204
134 142 147 153 136 140 142 121 131 80 62 68 42 64 56 57 74 96 121 131
216
192
186
190
171
163
127
162
177
149
167
214
207
216
230
218
202
203
188
152 138 153 142 151 136 113 120 82 38 31 36 33 39 36 59 91 90 101 125
216
151205
166210
154206
158197
145195
107192
1122016819949200362064122287228
150224
13022135233432274521586200
120 109
223
140227
140211
161218
164204
134213
1312159622857230412202322934236
191227
230221
224213
101226
12922170218
10620797 135
208
9 227
1973233
1943216
1246228
2147237
20
10
12
2396225
27
21
18
12
132215
137214
141212
147227
130221
116222
114
172236
226218
218224
200219
137212
109211
121 27
145
190
201
215
205
227
222
225
230
225
218
222
222
228
224
223
217
216
214
200
17
16
13
17
19
17
21
23
17
32
26
22
13
138 146 149 160 130 114 118 76 43 40 48 108 150 99 35 47 37 73 115 1698
133 116 147 115 96 134 24
118 2471 2421 21 9 2924 2481 15
159 1650 940 2434 2460 15
122 889 17
113
9
117 137 131 135 129 110 15
116 1177 1633 1416 1125 1638 935 1145 1237 1253 20
117 11
105 10
139 147
6 18
127 134 143 161 147 140 20
132 17
102 1394 954 1551 1654 1645 1049 1063 14
112 16
124 134
150 23
157
8 16
129 132 150 122 137 135 10
108 18
120 14
112 12
131 1299 13
102 130
116 23
116 24
111 27
142 28
138 33
148 37
152
13
16
9
16
17
17
9
9
34
28
21
75
120
43
105 135 123 133 137 137 138 128 132 127 135 131 140 147 119 142 142 146 151 138
7 13
9 23
58 107 128 132 139 151 18
140 18
144 14
156 142
149 17
145 156
146 36
136 19
155 41
161177
158201
152143
148
124 120 135 132 121 130 25
132 23
150 22
147 17
164 19
156 22
136 30
146 26
141 31
129 34
152 44
155157
159194
150150
141
140 151 119 122 115 118 19
132 25
128 19
144 17
152 19
156 16
140 17
148 24
142 28
121 26
151 26
150 65
158105
157 64
140
19 24 19 13 12 13
9 16 21 12 17 37 52 20
23 28 17 16 13 16 25 17 17 20 16 21 27 27
23 23 19 12
9 20 24 19 20 26 32 32 27 28
36 22 16 25 11 14 21 20 12
9 15 18 10
4
12 12 18 13
8
8
8 19 16
5
8 23
4
3
27 20 19 22 17 20 18 26 21 16 18 14
8 15
28 24 26 18 15 13 22 22 16 16 14 13 10 15
21 21 22 27 22 17 20 13 16 11 20 14
8 10
RGB
185 186 187
84 110 91
129 94 114
139 122 147
139 146 135
153 146 147
129 140 143
132 145 135
135 123 143
25
143 17
150 32
153
21
136 27
143 25
142
17
141 44
147 27
147
18
153 28
152 34
150
25
165 34
149 32
144
38
136 37
135 33
133
24
147 28
146 38
145
43
137126
123 76
129
70
150187
144128
131
18
141 61
125 52
141
15 34 20
26 27 17
39 18 11
7
7
3
6
2
5
22
5 13
9 10 11
10 14
6
188
119
128
125
119
150
138
124
133
27
143
22
146
14
143
24
136
18
133
41
122
36
135
26
133
36
141
20
141
13
26
17
18
4
10
5
5
33
29
30
33
15
37
26
49
58
28
23
27
15
11
4
4
13
4
28
31
32
22
15
37
30
34
49
31
17
16
8
5
5
16
14
7
42
23
19
18
20
31
21
20
34
13
12
21
12
6
3
5
2
10
33
24
22
18
15
23
18
25
30
24
17
22
14
14
3
8
8
8
30
18
14
11
14
19
18
25
25
18
21
29
23
16
15
16
10
10
21
15
15
17
16
28
36
35
15
7
11
28
30
26
21
16
18
12
7
Red-Green-Blue Matrices
Introduction to Image Processing
Red-Green-Blue Color Cube
R = 0 to 255
G = 0 to 255
Blue
B = 0 to 255
255
White
Cyan
Not all colors
are perceivable
on all devices
Magenta
255, 255, 255
Grey Line
Black
255
0, 0, 0
Green 255
Yellow
24-bit graphics:
256 x 256 x 256 = 16,777,216 colors
256 shades of grey
Red
0.0 to 1.0
0 to 255 (8-bits)
8
4
Introduction to Image Processing
Pixels: Color
8 bits
8 bits
8 bits
Red
Green
Blue
24 bits
RGB
9
Introduction to Image Processing
Pixel Profile of RGB Image
B
A
10
5
Introduction to Image Processing
Pixel Profile of RGB Image
A
B
From www.efg2.com/Lab/ImageProcessing/PixelProfile.htm
11
Introduction to Image Processing
• Pixels: Grey Scale
• Pixels: Color
• Color Issues
• Image Manipulation
• Image Enhancement
12
6
Introduction to Image Processing
Color Issue: Too Many Bits
Pixels with Axon Scanner
16 bits Red (Cy5)
16 bits Green (Cy3)
1111 0010 1101 0100
24-bit RGB Pixel:
1010 0101 1010 0101
1111 0010 1010 0101 0000 0000
Similar problem with 14-bit microscope images displayed in color.
Analysis of raw data may show details not seen on an image.
13
Introduction to Image Processing
Color Issue: Fidelity
CRT
Gamut
From www.efg2.com/Lab/Graphics/Colors/Chromaticity.htm
14
7
Introduction to Image Processing
3D Color Gamut
From “Visualization of Expanded Printing Gamuts Using 3-Dimensional Convex Hulls”
by Karl Guyler, Hallmark Cards, Kansas City
Note: Color calibration can be used to minimize needless differences
15
Introduction to Image Processing
• Pixels: Grey Scale
• Pixels: Color
• Color Issues
• Image Manipulation
– Image Registration
– Image Comparison
• Image Enhancement
16
8
Introduction to Image Processing
Image Manipulation
Comparison without Registration
17
Introduction to Image Processing
Image Manipulation: Registration
Two registration points needed
if image has rotation
18
9
Introduction to Image Processing
Image Manipulation: Registration
Translation and Rotation:
Align second image with first
19
Introduction to Image Processing
Image Manipulation: Comparison
Comparison with Alignment
20
10
Image Processing at CooperSurgical
Aceto-Positive Cervical Lesion Research
Original images from the web site of the
American Society for Colposcopy and Cervical Pathology
www.asccp.org
21
Introduction to Image Processing
•
•
•
•
•
Pixels: Grey Scale
Pixels: Color
Color Issues
Image Manipulation
Image Enhancement
– Contrast Improvement
– Spatial Filters
22
11
Introduction to Image Processing
Image Enhancement: Contrast Improvement
Original Image
2
3
2
3
2
4
Enhanced Image
3
4
3
1
3
1
6
6
5
5
4
4
Stretch
2
3
5
3
Histostretched
Original Histogram
3
3
1
5
1
3
2
1
0
1
2
3
4
5
to improve
contrast
0
1
2
3
4
5
23
Introduction to Image Processing
Image Enhancement: Contrast Improvement
From www.efg2.com/Lab/ImageProcessing/HistoStretchGrays.htm
24
12
Introduction to Image Processing
Image Enhancement: Contrast Improvement
25
Introduction to Image Processing
Image Enhancement: Contrast Improvement
Histogram stretching the color planes separately will yield an image with maximum contrast, but may
introduce a color shift. The color shift here is negligible.
Usually, convert Red-Green-Blue (RGB) → Hue-Saturation-Intensity (HSI), stretch I to yield I’, and
26
then convert HSI’ → R’G’B’
13
Image Processing of Microarray Slides
Axon Sample: Human2 Preview
Original
Autoscale
27
Introduction to Image Processing
Spatial Frequency
Frequency = 1
1 Cycle
Frequency = 2
2 Cycles
28
14
Introduction to Image Processing
Spatial Filters: Convolution
Output Image
Input Image
Column
Column
Row
Convolution Kernel,
a smoothing kernel in
this example
106 101
102 82

 68 46
78 
47 
38 
1 1 1
1
1 1 1
9
1 1 1
Row
Single Pixel
106 + 101 + 78 + 102 + 82 + 47 + 68 + 46 + 38
= 74.2 ≈ 74
9
∑=1
29
Introduction to Image Processing
Spatial Filters: Convolution
Type
Low-Pass
Spatial Filter
Smooth
High-Boost
Spatial Filter
Sharpen
Laplace Filter
Edge Enhance
1 1 1
1
1 1 1
9
1 1 1
∑=1
− 1 − 1 − 1 ∑ = 1
− 1
9 − 1

− 1 − 1 − 1
1 0
0
1 − 4 1 


0
1 0
∑= 0
30
15
Introduction to Image Processing
Spatial Filter: Smooth
Original
Enhanced
31
Introduction to Image Processing
Spatial Filter: Sharpen (original)
Picture Copyright © Adaptec, Used with Permission
32
16
Introduction to Image Processing
Spatial Filter: Sharpen (enhanced)
33
Introduction to Image Processing
Spatial Filter: Edge Enhance
Original
Laplace Filter Mask
Enhanced
50%
34
17
Introduction to Image Processing
• Pixels: Grey Scale
• Pixels: Color
• Color Issues
• Image Manipulation
• Image Enhancement
35
Image Processing of Microarray Slides
Axon Example: Single Feature
36
18
Acknowledgements
Bioinformatics
Arcady Mushegian
CooperSurgical
Kerry Blair
Genomics
Ranjan Perera
Angel McKee
Diane Stark
George Washington University
Medical Center
John L. Marlow, MD
37
19