The 'Skin Segmentation Dataset' contains samples of skin and non-skin pixel colours for automated face detection in colour images. Table 1 shows only a small subset of these samples. (a) Using the k-NN classification approach (for k=3, and k =7) determine whether a pixel with colour values R=145. G=140, and B=122 should be categorised as a skin

icon
Related questions
Question

Hello, could you please show working out how to tackle this question. It is a question to answer using paper method. My end of year exam for Image Analysis is nearing and I have been informed questions like these will pop up. 

 

Thanks

The 'Skin Segmentation Dataset' contains samples of skin and non-skin pixel colours for
automated face detection in colour images. Table 1 shows only a small subset of these
samples.
(a)
(b)
Using the k-NN classification approach (for k=3, and k =7) determine whether a
pixel with colour values R=145, G=140, and B=122 should be categorised as a skin
pixel or not. Use the 'city-block distance' (aka 'Manhattan distance' or 'taxicab
geometry') metric for your calculation.
What would be the classification if the Nearest Mean Classifier (NMC) were used?
[Use the 'city-block distance' metric.]
(c) Cosine-similarity measure between two vectors A and B is defined by
Transcribed Image Text:The 'Skin Segmentation Dataset' contains samples of skin and non-skin pixel colours for automated face detection in colour images. Table 1 shows only a small subset of these samples. (a) (b) Using the k-NN classification approach (for k=3, and k =7) determine whether a pixel with colour values R=145, G=140, and B=122 should be categorised as a skin pixel or not. Use the 'city-block distance' (aka 'Manhattan distance' or 'taxicab geometry') metric for your calculation. What would be the classification if the Nearest Mean Classifier (NMC) were used? [Use the 'city-block distance' metric.] (c) Cosine-similarity measure between two vectors A and B is defined by
similarity = cos(0) =
139
220
159
223
253
228
232
209
148
223
234
228
251
140
223
225
217
215
149
157
210
198
254
234
240
Skin samples (C1)
Red (R) Green (G) Blue (B)
where A, and B; are components of vector A and B respectively.
[Note: distance = (1 - similarity).]
Determine the classification result if the cosine-similarity metric is used with the
NMC scheme for the skin-colour data.
76
142
98
153
195
167
186
146
97
149
187
172
211
93
158
154
158
139
103
105
153
124
190
190
187
A.B
||A||||B||
59
104
51
104
158
120
171
113
54
100
171
123
201
51
126
102
124
87
51
48
123
61
152
155
171
Table 1
71
ΣΑ, Β,
i=1
ΣΑΣ Β
i=1
3
84
121
152
112
93
144
88
93
250
19
129
102
201
32
61
88
11
145
104
254
135
100
Non-Skin samples (C2)
Red (R) Green (G) Blue (B)
14
68
172
193
158
147
163
127
185
136
139
247
29
174
132
54
82
107
156
31
186
124
193
180
127
198
175
16
115
173
195
156
149
187
150
137
242
98
177
0
60
81
105
227
32
188
249
14
185
32
200
177
Transcribed Image Text:similarity = cos(0) = 139 220 159 223 253 228 232 209 148 223 234 228 251 140 223 225 217 215 149 157 210 198 254 234 240 Skin samples (C1) Red (R) Green (G) Blue (B) where A, and B; are components of vector A and B respectively. [Note: distance = (1 - similarity).] Determine the classification result if the cosine-similarity metric is used with the NMC scheme for the skin-colour data. 76 142 98 153 195 167 186 146 97 149 187 172 211 93 158 154 158 139 103 105 153 124 190 190 187 A.B ||A||||B|| 59 104 51 104 158 120 171 113 54 100 171 123 201 51 126 102 124 87 51 48 123 61 152 155 171 Table 1 71 ΣΑ, Β, i=1 ΣΑΣ Β i=1 3 84 121 152 112 93 144 88 93 250 19 129 102 201 32 61 88 11 145 104 254 135 100 Non-Skin samples (C2) Red (R) Green (G) Blue (B) 14 68 172 193 158 147 163 127 185 136 139 247 29 174 132 54 82 107 156 31 186 124 193 180 127 198 175 16 115 173 195 156 149 187 150 137 242 98 177 0 60 81 105 227 32 188 249 14 185 32 200 177
Expert Solution
steps

Step by step

Solved in 6 steps

Blurred answer