1
00:00:00,000 --> 00:00:02,548
Great. So let's go ahead and play around

2
00:00:02,549 --> 00:00:05,819
with OpenCV a little bit, and specifically,

3
00:00:05,819 --> 00:00:07,769
what you'll learn in this lecture is

4
00:00:07,799 --> 00:00:10,499
you will learn how to load images in

5
00:00:10,499 --> 00:00:13,259
Python using OpenCV, and you will learn

6
00:00:13,259 --> 00:00:16,199
how to display them, resize, and then

7
00:00:16,199 --> 00:00:20,121
save the resize images in a new image files.

8
00:00:20,541 --> 00:00:22,384
So load, display, resize, and

9
00:00:22,409 --> 00:00:26,789
write images, and I've got a nice image

10
00:00:26,789 --> 00:00:30,299
here of a galaxy. So I'll play around

11
00:00:30,299 --> 00:00:32,969
with this. The first thing you want to

12
00:00:32,969 --> 00:00:36,779
do is import library, and that the second

13
00:00:36,779 --> 00:00:38,669
thing is you want to load the image in

14
00:00:38,669 --> 00:00:42,642
Python. So image would be equal to

15
00:00:42,667 --> 00:00:48,214
cv2.imageread method. So the

16
00:00:48,239 --> 00:00:50,879
method expects now the path to the image

17
00:00:50,879 --> 00:00:53,159
that you want to load in Python, and

18
00:00:53,159 --> 00:00:57,839
that would be galaxy.jpg. So my

19
00:00:57,839 --> 00:01:00,449
script1.py file is in the

20
00:01:00,449 --> 00:01:04,109
same directory with galaxy.jpg,

21
00:01:04,349 --> 00:01:06,569
so I just need to pass the file name

22
00:01:06,594 --> 00:01:10,164
here, and then there is yet another parameter

23
00:01:10,384 --> 00:01:12,754
for imageread, for imread

24
00:01:12,784 --> 00:01:16,984
method, and then this parameter takes three

25
00:01:17,009 --> 00:01:19,949
arguments. Now this parameter specifies

26
00:01:19,949 --> 00:01:21,929
how you want to read the image in

27
00:01:21,964 --> 00:01:24,904
Python. So do you want to read it as an

28
00:01:24,932 --> 00:01:27,689
RBG image which means do you want three

29
00:01:27,689 --> 00:01:30,539
bands in your image, so you want a color

30
00:01:30,539 --> 00:01:33,779
image with a red band, the blue, and

31
00:01:33,779 --> 00:01:36,209
green band, so if you want to read the

32
00:01:36,209 --> 00:01:38,999
image as it is, so with colors, you'd

33
00:01:38,999 --> 00:01:40,197
want to pass 1 here.

34
00:01:40,222 --> 00:01:42,227
[No audio]

35
00:01:42,251 --> 00:01:45,659
If you want to read the image as black and white

36
00:01:45,659 --> 00:01:48,479
image in a grayscale, you'd want to pass

37
00:01:48,479 --> 00:01:51,449
0, and having a grayscale image

38
00:01:51,449 --> 00:01:53,849
implies that your image will have one

39
00:01:53,849 --> 00:01:56,159
band, and I'll get back to bands and

40
00:01:56,159 --> 00:01:59,429
explain them in just a moment. So and we

41
00:01:59,429 --> 00:02:01,649
also have -1, and this means the

42
00:02:01,649 --> 00:02:05,369
color image, but you also have an alpha

43
00:02:05,369 --> 00:02:07,769
channel, which means your image will

44
00:02:07,769 --> 00:02:10,709
have transparency capabilities. So if

45
00:02:10,709 --> 00:02:12,629
you apply operations that require

46
00:02:12,629 --> 00:02:16,319
transparency, you can do that when you

47
00:02:16,319 --> 00:02:18,119
read, when you load the image with a

48
00:02:18,119 --> 00:02:20,699
-1 argument here. Okay, so

49
00:02:20,699 --> 00:02:22,729
I would like to try out 0.

50
00:02:22,753 --> 00:02:24,924
[No audio]

51
00:02:24,949 --> 00:02:27,539
Great. Now, before I show the image, before I

52
00:02:27,539 --> 00:02:29,669
display the image on the screen, I'd

53
00:02:29,669 --> 00:02:33,569
like you to understand what this image

54
00:02:33,599 --> 00:02:37,618
object is about. So I'd want to print the type of it

55
00:02:37,642 --> 00:02:39,777
[No audio]

56
00:02:39,802 --> 00:02:40,811
just like that,

57
00:02:40,836 --> 00:02:45,232
[No audio]

58
00:02:45,316 --> 00:02:49,355
execute the script, script1. Yeah,

59
00:02:50,031 --> 00:02:52,899
close and then try again.

60
00:02:52,924 --> 00:02:54,934
[No audio]

61
00:02:54,959 --> 00:02:58,732
So this is a NumPy n-dimensional array,

62
00:02:58,756 --> 00:03:00,814
[No audio]

63
00:03:00,839 --> 00:03:02,688
and if you want, you can print that out,

64
00:03:02,712 --> 00:03:06,868
[No audio]

65
00:03:06,893 --> 00:03:09,103
and you'll see the actual NumPy array.

66
00:03:09,127 --> 00:03:11,298
[No audio]

67
00:03:11,323 --> 00:03:15,749
So this is a 2-dimensional array with values in the

68
00:03:15,790 --> 00:03:18,659
horizontal axis, and the vertical axis

69
00:03:18,659 --> 00:03:21,809
as well. So think of the image now and

70
00:03:21,839 --> 00:03:24,959
this would be the very first so the top

71
00:03:24,959 --> 00:03:28,019
left intensity value of the first pixel.

72
00:03:28,289 --> 00:03:31,199
So 14, for example, would be the

73
00:03:31,199 --> 00:03:34,079
intensity value in the grayscale, for

74
00:03:34,079 --> 00:03:36,539
the first pixel of the image, so for the

75
00:03:36,569 --> 00:03:39,929
top left pixel of the image, and then

76
00:03:39,929 --> 00:03:42,509
for the second pixel, and so on, and

77
00:03:42,509 --> 00:03:44,129
then you have these dots, which means

78
00:03:44,129 --> 00:03:46,679
Python cannot display the list, the

79
00:03:46,679 --> 00:03:48,749
long list in here because you have a

80
00:03:48,749 --> 00:03:51,179
couple of thousands values in the first

81
00:03:51,179 --> 00:03:52,949
row, and then you have the second row of

82
00:03:52,949 --> 00:03:55,799
pixels in the image, and the third, and so

83
00:03:55,799 --> 00:03:58,829
on, and this maze of metrics of the

84
00:03:58,829 --> 00:04:00,929
pixels of the image, and if you want to

85
00:04:00,929 --> 00:04:03,179
know how many numbers, how many values

86
00:04:03,179 --> 00:04:05,009
you have in the horizontal direction,

87
00:04:05,009 --> 00:04:06,779
how many values you have in the vertical

88
00:04:06,804 --> 00:04:09,504
directions, you can go ahead and print

89
00:04:10,114 --> 00:04:15,844
image.shape. Okay. So let's say that the

90
00:04:15,869 --> 00:04:20,855
image resolution is 1485 by 990.

91
00:04:22,322 --> 00:04:26,669
So Python stores the image as a NumPy array,

92
00:04:27,599 --> 00:04:30,359
as a matrix of numbers. As easy as that,

93
00:04:31,084 --> 00:04:33,574
if you want to check the dimensionals of

94
00:04:33,599 --> 00:04:37,319
your array of your image, you can do that

95
00:04:37,319 --> 00:04:40,619
with this expression, and you see that

96
00:04:40,619 --> 00:04:43,679
you have two dimensions. Now, if this

97
00:04:43,679 --> 00:04:47,549
was a color image, so with three bands,

98
00:04:47,699 --> 00:04:49,196
red, blue, and green,

99
00:04:49,220 --> 00:04:51,491
[No audio]

100
00:04:51,516 --> 00:04:52,872
things will change a little bit.

101
00:04:53,272 --> 00:04:55,732
So you've got 3-dimensions,

102
00:04:55,756 --> 00:04:57,987
[No audio]

103
00:04:58,012 --> 00:04:59,669
and you also see that the

104
00:04:59,669 --> 00:05:03,299
new array is a bit different. So here

105
00:05:03,299 --> 00:05:05,819
you've got values for each of the bands.

106
00:05:05,819 --> 00:05:08,519
So for green, for red, green, and blue.

107
00:05:10,079 --> 00:05:13,199
So I'd like to stick with a great image.

108
00:05:13,889 --> 00:05:16,589
And what I can do now is, I can display

109
00:05:16,589 --> 00:05:19,619
the image on the screen, and for that, you

110
00:05:19,619 --> 00:05:22,289
want to use the image. So imshow

111
00:05:22,319 --> 00:05:25,589
method, and this will display a window

112
00:05:25,589 --> 00:05:27,449
and you want to name that window. So you

113
00:05:27,449 --> 00:05:29,279
want to put a title for that window,

114
00:05:29,999 --> 00:05:31,121
let's say galaxy.

115
00:05:31,145 --> 00:05:33,574
[No audio]

116
00:05:33,599 --> 00:05:35,189
And what you pass here is

117
00:05:35,189 --> 00:05:39,449
the image object. So this one, great,

118
00:05:39,474 --> 00:05:42,388
and then what you want to do is

119
00:05:42,412 --> 00:05:44,802
[No audio]

120
00:05:44,827 --> 00:05:48,449
you want to specify a time for the audio

121
00:05:48,449 --> 00:05:51,959
window to be closed. Because this will

122
00:05:51,959 --> 00:05:54,239
show the window, but you also want to

123
00:05:54,239 --> 00:05:56,759
define some functionalities, so that the

124
00:05:56,759 --> 00:05:59,849
user can close the window. If you put 00

125
00:05:59,849 --> 00:06:02,129
here, and while the user presses any

126
00:06:02,129 --> 00:06:05,369
button, the window will close. Let me

127
00:06:05,369 --> 00:06:08,189
change this cv2. So if you put

128
00:06:08,189 --> 00:06:10,289
0 here, the user can close the window

129
00:06:10,289 --> 00:06:13,679
pressing any button. If you want to put

130
00:06:13,709 --> 00:06:16,859
a time, you could say 2000, and that

131
00:06:16,859 --> 00:06:21,449
simplized 2000 milliseconds. So that

132
00:06:21,449 --> 00:06:24,299
means 2 seconds. So we said how the

133
00:06:24,299 --> 00:06:26,729
user wants to close the window, and then

134
00:06:26,729 --> 00:06:29,849
you want to specify what to do when the

135
00:06:29,849 --> 00:06:33,119
user presses a button or waits for 2

136
00:06:33,119 --> 00:06:35,459
seconds. So he want to destroy all

137
00:06:35,489 --> 00:06:39,509
windows. That's the method that closes

138
00:06:39,509 --> 00:06:43,889
the window. Good. Let's see what

139
00:06:43,889 --> 00:06:48,188
will happen. Okay, so we got the image slide,

140
00:06:48,212 --> 00:06:51,020
and it waited for 2 seconds, and then it closed.

141
00:06:51,044 --> 00:06:53,480
[No audio]

142
00:06:53,505 --> 00:06:55,439
If you put it at 0,

143
00:06:55,463 --> 00:06:57,484
[No audio]

144
00:06:57,509 --> 00:07:01,049
the image will stay there, and if you

145
00:07:01,049 --> 00:07:04,739
press a button, it will close, and let me

146
00:07:04,739 --> 00:07:07,319
show the image again. Now, the reason

147
00:07:07,319 --> 00:07:10,199
that you don't see the image fit on my

148
00:07:10,199 --> 00:07:13,559
screen is that the image, as you see

149
00:07:13,559 --> 00:07:18,749
here in these values is 1485 pixels

150
00:07:18,779 --> 00:07:20,729
high. So the height, this is the

151
00:07:20,729 --> 00:07:24,869
height, and it's 990 pixels wide.

152
00:07:25,504 --> 00:07:27,304
So my screen resolution is set

153
00:07:27,329 --> 00:07:33,839
at 1280 by 720. So that means this image

154
00:07:33,839 --> 00:07:35,789
with this size does not fit on my

155
00:07:35,789 --> 00:07:38,055
screen, because my screen is too small for this.

156
00:07:38,080 --> 00:07:40,480
[No audio]

157
00:07:40,505 --> 00:07:42,288
So in this case, let me close this.

158
00:07:43,672 --> 00:07:45,604
What you can do is you want

159
00:07:45,629 --> 00:07:48,239
to resize the image, and then show the

160
00:07:48,239 --> 00:07:51,269
resized image. So we load the image,

161
00:07:51,299 --> 00:07:53,399
and before showing it, before passing it

162
00:07:53,431 --> 00:07:56,851
to the imshow method, you want to say

163
00:07:57,831 --> 00:07:59,526
let's say resized

164
00:07:59,558 --> 00:08:01,558
[No audio]

165
00:08:01,574 --> 00:08:03,509
image that will be equal

166
00:08:03,509 --> 00:08:09,089
to cv2.resize, and this would get

167
00:08:09,119 --> 00:08:11,339
two parameters. So the first is, of

168
00:08:11,339 --> 00:08:13,979
course, is the image object that you want

169
00:08:13,979 --> 00:08:17,969
to resize. So img is our variable, and

170
00:08:17,969 --> 00:08:20,069
then you want to specify a tuple with a

171
00:08:20,069 --> 00:08:25,049
new dimensions, I'll say 1000 by 500.

172
00:08:26,434 --> 00:08:29,552
And then you want to pass a new image here.

173
00:08:29,576 --> 00:08:33,538
[No audio]

174
00:08:33,563 --> 00:08:35,159
So what's happening here is that

175
00:08:35,969 --> 00:08:38,489
Python is actually resizing the NumPy

176
00:08:38,489 --> 00:08:41,429
array. So it will take the array with

177
00:08:41,429 --> 00:08:44,939
this number of pixels, number of values.

178
00:08:45,299 --> 00:08:47,609
and it will create an array with this

179
00:08:47,669 --> 00:08:50,399
new dimensions. So what will happen

180
00:08:50,399 --> 00:08:52,409
there is that Python will interpolate

181
00:08:52,409 --> 00:08:55,019
those values. So it has quite a lot of

182
00:08:55,019 --> 00:08:57,959
values here. But then it goes from this

183
00:08:58,859 --> 00:09:02,819
to this. So when it sees let's say it

184
00:09:02,819 --> 00:09:06,179
has 4 for a value and the 6

185
00:09:06,179 --> 00:09:09,689
for the neighbor value, and what Python

186
00:09:09,689 --> 00:09:13,636
will do is that it will get a 4 and 6

187
00:09:13,661 --> 00:09:16,379
basically and it will just make a

188
00:09:16,379 --> 00:09:18,419
value out of it, so let's say 5.

189
00:09:19,789 --> 00:09:22,220
That's basically the idea. So it terminates

190
00:09:22,245 --> 00:09:24,449
the values and then it shows

191
00:09:24,449 --> 00:09:26,369
the interpolated image on the screen,

192
00:09:27,064 --> 00:09:29,494
which looks pretty nice in our

193
00:09:29,519 --> 00:09:33,319
eyes. Okay, let's see this,

194
00:09:33,343 --> 00:09:36,519
[No audio]

195
00:09:36,544 --> 00:09:39,989
and this is the image and here you see that this is quite

196
00:09:39,989 --> 00:09:42,929
stretched a little bit. So it was a

197
00:09:42,954 --> 00:09:45,324
total image but now it's quite wide.

198
00:09:46,420 --> 00:09:50,310
And because this is actually the width

199
00:09:50,310 --> 00:09:52,320
of the image and this is the height, so

200
00:09:52,371 --> 00:09:57,501
if you want you can say 500 and 1000, so if

201
00:09:58,590 --> 00:10:02,670
again, and now you see more or less the two

202
00:10:02,700 --> 00:10:06,270
ratio of the image. But if you want

203
00:10:06,270 --> 00:10:09,480
to keep the ratio of the image, you'd

204
00:10:09,480 --> 00:10:13,200
want to go more advanced here. So let's

205
00:10:13,200 --> 00:10:16,980
say from this, so let's say we would

206
00:10:16,980 --> 00:10:19,650
want to show a size that is half of

207
00:10:19,650 --> 00:10:22,920
this. So that that would keep the ratio

208
00:10:22,920 --> 00:10:25,320
of the image. So what we can do is we

209
00:10:25,320 --> 00:10:28,740
need access to these values, and we can

210
00:10:28,740 --> 00:10:31,230
grab those values from the shape

211
00:10:31,230 --> 00:10:33,450
method. So this produces a tuple with

212
00:10:33,450 --> 00:10:36,510
these two values, and then we go here

213
00:10:36,510 --> 00:10:41,622
and say, image.shape,

214
00:10:41,663 --> 00:10:43,663
[No audio]

215
00:10:43,688 --> 00:10:45,360
and this would be

216
00:10:45,385 --> 00:10:49,763
this value, so 990, which has an index of 1,

217
00:10:51,522 --> 00:10:54,035
and then we have again image.shape

218
00:10:55,935 --> 00:10:58,021
index of 0 for this number,

219
00:10:58,045 --> 00:11:00,045
[No audio]

220
00:11:01,408 --> 00:11:04,221
and then we want to divide this by 2.

221
00:11:04,245 --> 00:11:06,398
[No audio]

222
00:11:06,423 --> 00:11:08,700
Okay, and I expects to get an

223
00:11:08,700 --> 00:11:12,000
error from this. Let's see. So

224
00:11:12,030 --> 00:11:14,655
yeah, we've got an error. It says TypeError

225
00:11:14,679 --> 00:11:16,890
integer argument expected, got

226
00:11:16,890 --> 00:11:20,580
float. But here, what we're doing here

227
00:11:20,580 --> 00:11:24,570
is that when we divide this number by

228
00:11:24,570 --> 00:11:27,000
2, we will get a float. So we would

229
00:11:27,000 --> 00:11:31,288
get something like 742.5.

230
00:11:31,312 --> 00:11:34,242
[No audio]

231
00:11:34,267 --> 00:11:36,821
In that case, what we want to do is

232
00:11:36,845 --> 00:11:38,321
convert this to an integer.

233
00:11:38,345 --> 00:11:41,027
[No audio]

234
00:11:41,078 --> 00:11:46,916
So 742.5 for becomes 742 .

235
00:11:47,945 --> 00:11:51,155
Okay, and let's keep the consistency so

236
00:11:51,180 --> 00:11:53,455
integers for these as well.

237
00:11:53,479 --> 00:11:57,087
[No audio]

238
00:11:57,112 --> 00:11:58,318
And let's see.

239
00:11:59,498 --> 00:12:02,520
And we've got an invalid syntax here. So it

240
00:12:02,520 --> 00:12:06,330
points us to the line 11, somewhere in the

241
00:12:06,360 --> 00:12:08,250
beginning, which can be quite

242
00:12:08,250 --> 00:12:12,600
misleading. So you want to see before

243
00:12:12,600 --> 00:12:15,360
that line, which is here, and you can

244
00:12:15,360 --> 00:12:17,730
see that these brackets here closes

245
00:12:17,730 --> 00:12:20,910
here, so we need another bracket, which

246
00:12:20,910 --> 00:12:23,070
closes in the first bracket here. So

247
00:12:23,070 --> 00:12:24,943
save that, try again.

248
00:12:24,967 --> 00:12:26,967
[No audio]

249
00:12:26,990 --> 00:12:29,255
And this time the image looks good.

250
00:12:30,502 --> 00:12:31,770
So you will learn how

251
00:12:31,770 --> 00:12:34,680
to load an image in Python, and you'll

252
00:12:34,680 --> 00:12:37,170
learn how to resize an image, how to show

253
00:12:37,170 --> 00:12:40,350
an image on the screen, and now let's go

254
00:12:40,350 --> 00:12:43,230
ahead and write the resized_image in

255
00:12:43,230 --> 00:12:46,200
a new file. For that, you'd want to use

256
00:12:46,235 --> 00:12:49,521
the imwrite methods. So imagewrite

257
00:12:50,837 --> 00:12:52,260
and you want to give a name to

258
00:12:52,290 --> 00:12:57,763
the new image Galaxy, let's say, resized.jpg,

259
00:12:57,787 --> 00:12:59,787
[No audio]

260
00:12:59,812 --> 00:13:01,290
and then you pass the image

261
00:13:01,290 --> 00:13:03,630
object that you want to store in this

262
00:13:04,200 --> 00:13:07,650
file. So the comma goes also here,

263
00:13:08,160 --> 00:13:11,488
and the image you want to store is

264
00:13:11,512 --> 00:13:14,420
resized_image. That's it.

265
00:13:14,444 --> 00:13:16,882
[No audio]

266
00:13:16,907 --> 00:13:19,680
Execute, and we got

267
00:13:19,710 --> 00:13:23,880
the old image resize on the fly. So

268
00:13:23,910 --> 00:13:26,550
Python gets a NumPy array, it

269
00:13:26,550 --> 00:13:29,550
interpolates it, so it resizes it and then

270
00:13:29,550 --> 00:13:31,800
it shows it on the screen. So this is

271
00:13:31,800 --> 00:13:35,490
galaxy, this one here, and then we can close

272
00:13:35,490 --> 00:13:39,150
this, and here we've got our new image.

273
00:13:39,390 --> 00:13:40,736
So Galaxy_resized.

274
00:13:40,761 --> 00:13:43,830
[No audio]

275
00:13:43,855 --> 00:13:44,940
So if you go to the

276
00:13:44,940 --> 00:13:48,270
folder where these files are, you'll see

277
00:13:48,270 --> 00:13:51,150
that this image has new dimensions. So

278
00:13:51,150 --> 00:13:56,010
495 by 742, and that's what I wanted to

279
00:13:56,010 --> 00:13:58,230
teach you in this lecture. Hope you

280
00:13:58,230 --> 00:13:59,940
enjoyed it and talk to you later.