1
00:00:05,450 --> 00:00:10,230
First, let's import the PIL
library and the image object.

2
00:00:10,230 --> 00:00:14,370
Import PIL then from
pill we'll import image.

3
00:00:14,370 --> 00:00:17,730
Let's import the display
functionality from

4
00:00:17,730 --> 00:00:22,740
IPython so from IPython.display
will import display.

5
00:00:22,740 --> 00:00:24,780
Finally, let's load the image

6
00:00:24,780 --> 00:00:26,775
we were working with last time.

7
00:00:26,775 --> 00:00:30,250
So that image is in
read_only/msi_recruitment.gif.

8
00:00:31,580 --> 00:00:34,770
I will open it into
the image object.

9
00:00:34,770 --> 00:00:36,705
Let's execute that so.

10
00:00:36,705 --> 00:00:38,730
Great. Now, let's check out

11
00:00:38,730 --> 00:00:40,980
a few more methods of
the image library.

12
00:00:40,980 --> 00:00:42,690
First, we're going
to look at copy.

13
00:00:42,690 --> 00:00:44,660
If you remember, we
can do this using

14
00:00:44,660 --> 00:00:47,345
the built-in Python help command.

15
00:00:47,345 --> 00:00:50,900
So help image.copy. All right.

16
00:00:50,900 --> 00:00:53,540
Here's some information
about copy.

17
00:00:53,540 --> 00:00:56,390
We see that the copy takes
no arguments and that

18
00:00:56,390 --> 00:00:59,360
the return object is
an image object itself.

19
00:00:59,360 --> 00:01:01,145
Now, let's look at save.

20
00:01:01,145 --> 00:01:03,755
So help image.save.

21
00:01:03,755 --> 00:01:08,480
Okay. There's quite a bit
to the image.save object.

22
00:01:08,480 --> 00:01:10,460
The image save method has

23
00:01:10,460 --> 00:01:12,440
a couple of parameters
which are interesting.

24
00:01:12,440 --> 00:01:14,510
The first called fp,

25
00:01:14,510 --> 00:01:17,315
is the filename we want
to save the object to.

26
00:01:17,315 --> 00:01:20,340
The second format is interesting.

27
00:01:20,340 --> 00:01:22,470
It allows us to
change the type of

28
00:01:22,470 --> 00:01:24,770
the image but the docs
tells us that this should

29
00:01:24,770 --> 00:01:26,600
be done automatically
by looking at

30
00:01:26,600 --> 00:01:29,960
the file extension as
well. Let's give it a try.

31
00:01:29,960 --> 00:01:32,860
This file was originally
a Gif image file,

32
00:01:32,860 --> 00:01:34,430
but I bet if we save it with

33
00:01:34,430 --> 00:01:36,935
a.png format and read it again,

34
00:01:36,935 --> 00:01:39,130
we'll get a different
kind of file.

35
00:01:39,130 --> 00:01:41,750
Let's image.save and
we'll just call it

36
00:01:41,750 --> 00:01:44,330
msi_recruitment.png
and that's going

37
00:01:44,330 --> 00:01:46,730
to save it in
your home directory.

38
00:01:46,730 --> 00:01:48,710
Then we'll open a new image.

39
00:01:48,710 --> 00:01:53,220
So image.open
msi_recruitment.png and

40
00:01:53,220 --> 00:01:54,935
take that into an image variable.

41
00:01:54,935 --> 00:01:57,080
Now, let's inspect like we did

42
00:01:57,080 --> 00:01:59,060
in the previous lecture and call

43
00:01:59,060 --> 00:02:04,765
inspect.getmro and wrap
image in the type function.

44
00:02:04,765 --> 00:02:07,820
Indeed, this created a new file

45
00:02:07,820 --> 00:02:09,110
which we could view by going into

46
00:02:09,110 --> 00:02:11,030
the Jupyter Notebook
file list by clicking

47
00:02:11,030 --> 00:02:13,220
on the logo at the top
of the browser,

48
00:02:13,220 --> 00:02:15,080
and we can see this new object is

49
00:02:15,080 --> 00:02:17,795
actually a png image file object.

50
00:02:17,795 --> 00:02:20,255
For the purposes of this class,

51
00:02:20,255 --> 00:02:21,980
the difference between
the image formats

52
00:02:21,980 --> 00:02:23,780
really isn't so important,

53
00:02:23,780 --> 00:02:26,720
but it's nice that you can
explore how a library works

54
00:02:26,720 --> 00:02:31,280
using the functions of
help, dir and getmro.

55
00:02:31,390 --> 00:02:33,980
The Pillow library also has

56
00:02:33,980 --> 00:02:37,160
some nice image filters
to add some effects.

57
00:02:37,160 --> 00:02:40,010
It does this through
the filter function.

58
00:02:40,010 --> 00:02:41,750
The filter function takes

59
00:02:41,750 --> 00:02:43,520
a filter object and those are all

60
00:02:43,520 --> 00:02:47,720
stored in the image filter
object. Let's take a look.

61
00:02:47,720 --> 00:02:49,910
So from PIL, we'll import

62
00:02:49,910 --> 00:02:51,575
the image filter object

63
00:02:51,575 --> 00:02:54,960
and then run help
on image filter.

64
00:02:56,120 --> 00:02:59,599
There are a bunch of
different filters here,

65
00:02:59,599 --> 00:03:02,795
but let's just try and
apply the blur filter.

66
00:03:02,795 --> 00:03:04,220
Before we do this,

67
00:03:04,220 --> 00:03:07,820
we have to convert the image
to what's called RGB mode.

68
00:03:07,820 --> 00:03:09,595
This is a bit magical.

69
00:03:09,595 --> 00:03:12,950
Images like Gifs are limited
into how many colors

70
00:03:12,950 --> 00:03:16,685
can be displayed at once based
on the size of the pallet.

71
00:03:16,685 --> 00:03:18,020
This is similar to

72
00:03:18,020 --> 00:03:21,500
a painter's palette which
only has so much room.

73
00:03:21,500 --> 00:03:24,605
This is actually
a very old image format.

74
00:03:24,605 --> 00:03:25,940
If we convert the image

75
00:03:25,940 --> 00:03:27,610
into something more
sophisticated,

76
00:03:27,610 --> 00:03:30,500
we can apply these
interesting image transforms.

77
00:03:30,500 --> 00:03:33,290
Sometimes, learning
a new library means digging a

78
00:03:33,290 --> 00:03:36,530
bit deeper into the domain
the library is about.

79
00:03:36,530 --> 00:03:40,280
We could convert this image
using the convert function.

80
00:03:40,280 --> 00:03:43,130
So image and then we'll just call

81
00:03:43,130 --> 00:03:46,010
image.convertRGB and
this stands for red,

82
00:03:46,010 --> 00:03:47,690
green and blue mode which

83
00:03:47,690 --> 00:03:50,050
is a pretty common
mode for images.

84
00:03:50,050 --> 00:03:52,130
Then I will create a new variable

85
00:03:52,130 --> 00:03:53,885
blurred image and
we'll set that to

86
00:03:53,885 --> 00:03:56,000
image.filter and we'll pass

87
00:03:56,000 --> 00:03:59,315
in the PIL.imagefilter.blur
parameter.

88
00:03:59,315 --> 00:04:01,610
You'll note that the parameter
that we're passing in

89
00:04:01,610 --> 00:04:03,950
is really a placeholder,

90
00:04:03,950 --> 00:04:05,810
it's not a function that

91
00:04:05,810 --> 00:04:08,185
we're running it's an
object we're passing in.

92
00:04:08,185 --> 00:04:11,770
Then, let's display
that blurred image.

93
00:04:12,400 --> 00:04:15,860
I encourage you to pause
the video here and jump into

94
00:04:15,860 --> 00:04:16,880
the notebooks and start

95
00:04:16,880 --> 00:04:19,655
experimenting with some
of the other filters.

96
00:04:19,655 --> 00:04:21,260
The emboss and sharpen

97
00:04:21,260 --> 00:04:23,525
filters for instance
are interesting.

98
00:04:23,525 --> 00:04:24,695
Or for a challenge,

99
00:04:24,695 --> 00:04:26,450
check out the box blur or

100
00:04:26,450 --> 00:04:28,370
the median filter
functions and look

101
00:04:28,370 --> 00:04:29,750
at their parameters to get

102
00:04:29,750 --> 00:04:33,030
a sense of how
they're being used.

103
00:04:35,210 --> 00:04:37,660
Okay. Let me show you

104
00:04:37,660 --> 00:04:40,555
one more function in
this lecture, which is crop.

105
00:04:40,555 --> 00:04:42,580
This removes portions of

106
00:04:42,580 --> 00:04:45,745
the image except for the
bounding box that you describe.

107
00:04:45,745 --> 00:04:49,060
When you think of images
think of individual dots or

108
00:04:49,060 --> 00:04:52,390
pixels which make up the image
being lined up on a grid.

109
00:04:52,390 --> 00:04:55,150
You can actually see
the number of pixels high

110
00:04:55,150 --> 00:04:58,240
the image is and
the width of the image.

111
00:04:58,240 --> 00:04:59,890
So let's do that here,

112
00:04:59,890 --> 00:05:01,930
we'll print the image.

113
00:05:01,930 --> 00:05:03,340
We'll use the string

114
00:05:03,340 --> 00:05:06,010
formatting to pass
in two parameters.

115
00:05:06,010 --> 00:05:08,410
The first is going
to be image.width,

116
00:05:08,410 --> 00:05:10,420
the width of the
image in pixels and

117
00:05:10,420 --> 00:05:14,060
image.height the height
of the image in pixels.

118
00:05:14,240 --> 00:05:18,390
We see it's 800 by 450.

119
00:05:18,390 --> 00:05:20,290
This means that this image is

120
00:05:20,290 --> 00:05:22,330
800 pixels wide and we call that

121
00:05:22,330 --> 00:05:25,090
the x-axis and 450 pixels

122
00:05:25,090 --> 00:05:27,440
high and we call that the y-axis.

123
00:05:27,440 --> 00:05:30,070
If we take a look at
the crop documentation,

124
00:05:30,070 --> 00:05:31,810
we see that the first parameter

125
00:05:31,810 --> 00:05:33,550
to the function is a tuple,

126
00:05:33,550 --> 00:05:35,140
which is the left upper,

127
00:05:35,140 --> 00:05:38,770
right and lower values
of the x, y coordinates.

128
00:05:38,770 --> 00:05:40,780
So the two corners of the image.

129
00:05:40,780 --> 00:05:44,110
So help image.crop, we run

130
00:05:44,110 --> 00:05:48,710
that and we see that there's
a box that we provide.

131
00:05:48,710 --> 00:05:51,519
With PIL images, we define

132
00:05:51,519 --> 00:05:52,930
the bounding box using

133
00:05:52,930 --> 00:05:55,600
the upper left corner
and lower right corner.

134
00:05:55,600 --> 00:05:57,970
We count the number
of pixels out from

135
00:05:57,970 --> 00:06:00,455
the upper left corner
which is zero zero.

136
00:06:00,455 --> 00:06:02,420
This might seem odd
if you're used to

137
00:06:02,420 --> 00:06:04,910
coordinate systems which
start in the lower left,

138
00:06:04,910 --> 00:06:08,990
a lot of us learned these in
primary school mathematics.

139
00:06:08,990 --> 00:06:11,630
Just remember that
we define our box in

140
00:06:11,630 --> 00:06:16,320
the same way we count out
the positions in the image.

141
00:06:16,330 --> 00:06:18,920
So if we wanted to get

142
00:06:18,920 --> 00:06:21,245
the Michigan logo
out of this image,

143
00:06:21,245 --> 00:06:22,805
we might start with the left,

144
00:06:22,805 --> 00:06:27,395
I'd say 50 pixels and
the top at zero pixels.

145
00:06:27,395 --> 00:06:30,650
Then we might walk to
the right another 190

146
00:06:30,650 --> 00:06:35,050
pixels and set the lower
bound to say a 150 pixels.

147
00:06:35,050 --> 00:06:38,840
Let's display and we'll
call image.crop and we

148
00:06:38,840 --> 00:06:42,065
pass it a parameter is a tuple.

149
00:06:42,065 --> 00:06:44,360
Remember that we're
going to put all four of

150
00:06:44,360 --> 00:06:46,640
these parameters in
one tuple object and

151
00:06:46,640 --> 00:06:48,470
pass it in and this
should display

152
00:06:48,470 --> 00:06:51,050
the image. There we go.

153
00:06:51,050 --> 00:06:52,250
That's the School of

154
00:06:52,250 --> 00:06:55,700
Information logo cropped
out of the image.

155
00:06:55,700 --> 00:06:59,300
Of course, crop like
other functions only returns

156
00:06:59,300 --> 00:07:03,455
a copy of the image and doesn't
change the image itself.

157
00:07:03,455 --> 00:07:06,380
A strategy I like to
do is to try and draw

158
00:07:06,380 --> 00:07:08,180
the bounding box directly on

159
00:07:08,180 --> 00:07:11,015
the image where I'm
trying to line things up.

160
00:07:11,015 --> 00:07:15,110
We can draw on images using
the image draw object.

161
00:07:15,110 --> 00:07:17,150
I'm not going to go
into this in detail,

162
00:07:17,150 --> 00:07:19,910
but here's a quick
example of how I

163
00:07:19,910 --> 00:07:23,345
might draw the bounding box
in this particular case.

164
00:07:23,345 --> 00:07:28,225
So from PIL we'll import
the image draw object,

165
00:07:28,225 --> 00:07:30,525
we'll instantiate
this, so we'll call

166
00:07:30,525 --> 00:07:32,805
imagedraw.draw and we
pass it the image.

167
00:07:32,805 --> 00:07:35,120
You can check the
documentation to learn more

168
00:07:35,120 --> 00:07:38,315
about how you would
use this object.

169
00:07:38,315 --> 00:07:40,190
Then we say drawing

170
00:07:40,190 --> 00:07:42,800
object.rectangle and we
give it the rectangle

171
00:07:42,800 --> 00:07:44,390
the outline that we're actually

172
00:07:44,390 --> 00:07:47,155
interested in and we
give it a fill value.

173
00:07:47,155 --> 00:07:49,250
You can set the fill
to different colors or

174
00:07:49,250 --> 00:07:51,320
types and an outline,

175
00:07:51,320 --> 00:07:53,930
in this case we want
to make it red.

176
00:07:53,930 --> 00:07:56,480
Because we ran this on
the drawing object,

177
00:07:56,480 --> 00:07:57,950
it actually executes it and

178
00:07:57,950 --> 00:07:59,540
the drawing object
has a reference

179
00:07:59,540 --> 00:08:00,950
to the underlying image.

180
00:08:00,950 --> 00:08:03,290
So let's display
this image now in

181
00:08:03,290 --> 00:08:10,385
the Jupyter Notebook.
There we go.

182
00:08:10,385 --> 00:08:14,015
We have an image which
is our main image of

183
00:08:14,015 --> 00:08:20,065
the logo that's outlined in
the red box for our flier.

184
00:08:20,065 --> 00:08:22,700
Okay. That's been
an overview of how

185
00:08:22,700 --> 00:08:24,785
to use PIL for single images,

186
00:08:24,785 --> 00:08:26,510
but a lot of work might involve

187
00:08:26,510 --> 00:08:28,400
multiple images and in fact,

188
00:08:28,400 --> 00:08:31,175
your assignment involves
multiple images.

189
00:08:31,175 --> 00:08:33,710
So in the next lecture, we're
going to tackle that and

190
00:08:33,710 --> 00:08:37,140
set you up for this assignment.
I'll see you there.