1 00:00:00,030 --> 00:00:05,790 Great, so let's go ahead and play around with OpenCV a little bit. Specifically 2 00:00:05,790 --> 00:00:10,920 what you'll learn in this lecture is you will learn how to load images in Python 3 00:00:10,920 --> 00:00:17,250 using OpenCV. You'll learn how to display them, resize, and then save the resized 4 00:00:17,250 --> 00:00:24,750 images in new images image files, so load display, resize, and write images. 5 00:00:24,750 --> 00:00:32,189 And I've got a nice image here of a galaxy, so I'll play around with this. The first 6 00:00:32,189 --> 00:00:37,559 thing you want to do is import the library, and then the second thing is you 7 00:00:37,559 --> 00:00:44,969 want to load the image in Python so image would be equal to cv2.image 8 00:00:44,969 --> 00:00:51,480 read method, so the method expects now the path to the image that you want to 9 00:00:51,480 --> 00:00:59,370 load in Python and that would be galaxy.jpg. So I, my script1.py file 10 00:00:59,370 --> 00:01:05,850 file is in the same directory with galaxy.jpg, so i just need to pass 11 00:01:05,850 --> 00:01:12,299 the file name here. And then there is yet another parameter for image read, 12 00:01:12,299 --> 00:01:19,170 for imread method and then this parameter takes three arguments. Now, this parameter 13 00:01:19,170 --> 00:01:24,299 specifies how you want to read the image in Python so do you want to read it as 14 00:01:24,299 --> 00:01:28,799 our RBG image which means you want three bands in your image, 15 00:01:28,799 --> 00:01:35,250 so you want a color image with a red band, blue, and green band. So if you 16 00:01:35,250 --> 00:01:42,299 want to read the image as it is, so with colors you'd want to pass one here. 17 00:01:42,299 --> 00:01:48,060 If you want to read the image as black and white image, so in a grayscale you'd want 18 00:01:48,060 --> 00:01:53,790 to pass zero, and having a grayscale image implies that your image will have one 19 00:01:53,790 --> 00:01:59,430 band and I'll get back to bands and explain them in just a moment. So we 20 00:01:59,430 --> 00:02:05,729 also have minus one. This means the color image, but you also have an alpha channel 21 00:02:05,729 --> 00:02:10,830 which means your image will have transparency capabilities, so if you 22 00:02:10,830 --> 00:02:14,170 apply operations that require transparency 23 00:02:14,170 --> 00:02:19,180 you can do that when you read, when you load the image with a minus one argument here. 24 00:02:19,180 --> 00:02:27,310 Okay, so I would like to try out 0. Great, now before I show the image, before 25 00:02:27,310 --> 00:02:33,520 I display the image on the screen I'll like you to understand what this image 26 00:02:33,520 --> 00:02:46,000 object is about, so I'd want to print the type of it, just like that. Execute the 27 00:02:46,000 --> 00:02:56,520 script, script1, yeah. Of course and try again, so this is a Numpy N 28 00:02:56,520 --> 00:03:07,660 dimensional array, and if you want you can print that out, and you'll see the 29 00:03:07,660 --> 00:03:15,510 actual Numpy array so this is a two-dimensional array with values in the 30 00:03:15,510 --> 00:03:21,520 horizontal axis and in a vertical axis as well. So think of the image now and 31 00:03:21,520 --> 00:03:28,570 this would be the very first, so the top left intensity value of the first pixel. 32 00:03:28,570 --> 00:03:34,540 So 14 for example would be the intensity value in the grayscale for the first 33 00:03:34,540 --> 00:03:40,540 pixel of image so for the top left pixel of the image, and then for the second 34 00:03:40,540 --> 00:03:45,150 pixel and so on and then you have these dots which means Python cannot display 35 00:03:45,150 --> 00:03:50,320 ever listed a long list in here because you have a couple of thousands of values 36 00:03:50,320 --> 00:03:54,070 in the first row, and then you have a second row of pixels in the image 37 00:03:54,070 --> 00:04:00,550 and the third and so on, and this makes the matrix of the pixel image, and if 38 00:04:00,550 --> 00:04:04,300 you want to know how many numbers how many values you have in the horizontal 39 00:04:04,300 --> 00:04:08,770 direction, how many values you have in the vertical direction you can go ahead 40 00:04:08,770 --> 00:04:19,500 and print image shape. Okay. So let's say that the image resolution 41 00:04:19,500 --> 00:04:26,030 is 1485 by 990. So Python stores the image as a Numpy 42 00:04:26,030 --> 00:04:32,180 array, as a matrix of numbers as easy as that. If you want to check the 43 00:04:32,180 --> 00:04:40,280 dimensional of your array, of your image you can do that with this expression and you 44 00:04:40,280 --> 00:04:47,120 see that you have two dimensions. Now if they sauce a color image so with three 45 00:04:47,120 --> 00:04:52,910 bands, red, blue, and green things will change a little bit. 46 00:04:52,910 --> 00:05:01,940 So you've got three dimensions and you also see that the new array is a bit 47 00:05:01,940 --> 00:05:07,280 different, so here you've got values for each of the band so for green, to red 48 00:05:07,280 --> 00:05:15,650 green, and blue, so I'd like to stick with the gray image, and what I can do now is 49 00:05:15,650 --> 00:05:21,530 I can display the image on the screen and for that you want to use the image so 50 00:05:21,530 --> 00:05:27,199 imshow method and this will display a window and you want to name that window, 51 00:05:27,199 --> 00:05:34,760 so you want to put a title for that window. Let's say galaxy. What you pass here 52 00:05:34,760 --> 00:05:45,380 is the image object, so this one. Great. And then what you want to do is you want 53 00:05:45,380 --> 00:05:52,430 to specify a time for your window to be closed because this will show the 54 00:05:52,430 --> 00:05:57,110 window but you also want to define some functionalities so that the user can 55 00:05:57,110 --> 00:06:02,900 close the window. If you put 0 here when the user presses any button, the window 56 00:06:02,900 --> 00:06:09,620 will close. Let me change this to cv2. So if you put 0, the user can 57 00:06:09,620 --> 00:06:16,280 close the window pressing any button. If you want to put a time you could say 58 00:06:16,280 --> 00:06:23,990 2 000 and that implies 2 000 milliseconds, so that means 2 seconds. So we said how 59 00:06:23,990 --> 00:06:29,690 the user wants to close the window and then you want to specify what to do when 60 00:06:29,690 --> 00:06:37,220 the user presses a button or waits for 2 seconds, so we want to destroy all windows. 61 00:06:37,220 --> 00:06:45,200 That's the method that closes the window. Good, let's see what will happen. 62 00:06:45,200 --> 00:06:50,490 Okay so we got the image displayed and it waited for two seconds and then it closed. 63 00:06:50,490 --> 00:07:01,410 If you put it at zero the image will stay there, and if you press the 64 00:07:01,410 --> 00:07:07,650 button it will close. And let me show the image again. Now the reason that you 65 00:07:07,650 --> 00:07:14,520 don't see the image fit on my screen is that the image as you see in these values 66 00:07:14,520 --> 00:07:24,540 is 1,485 pixels high so the height, this is the height and it's 990 pixels wide. 67 00:07:24,540 --> 00:07:33,720 So my screen resolution is set at 1280 by 720, so that means that this image 68 00:07:33,720 --> 00:07:39,620 with this size doesn't fit on my screen because my screen is too small for this. 69 00:07:39,620 --> 00:07:46,110 So in this case, let me close this. What you can do is you want to resize 70 00:07:46,110 --> 00:07:52,110 the image and then show the resized image. So we load the image and before showing 71 00:07:52,110 --> 00:08:00,830 it, before passing it to the imshow method you want to say let's say resized 72 00:08:00,830 --> 00:08:09,930 image. That will be equal to cv2.resize. And this would get two parameters, 73 00:08:09,930 --> 00:08:15,840 so the first is of course is the image object that you want to resize, so img is 74 00:08:15,840 --> 00:08:20,700 our variable, and then you want to specify a tuple with the new dimensions. 75 00:08:20,700 --> 00:08:34,050 I'll say 1,000 by 500 and then you want to pass the new image here, so what's 76 00:08:34,050 --> 00:08:39,930 happening here is that Python is actually resizing the Numpy array, 77 00:08:39,930 --> 00:08:45,720 so it will take the array with this number of pixels, number of values 78 00:08:45,720 --> 00:08:50,090 and it will create an array with these new dimensions, so what will 79 00:08:50,090 --> 00:08:54,650 happen there is that Python will interpolate those values, so it has quite 80 00:08:54,650 --> 00:09:02,000 a lot of values here but then it goes from this to this, so when it sees let's 81 00:09:02,000 --> 00:09:09,740 say it has 4 for one value and then six for the neighbor value, and what Python 82 00:09:09,740 --> 00:09:16,280 will do is that it will get 4 and 6 basically, and it will just make a value 83 00:09:16,280 --> 00:09:22,400 out of it, so let's say five, that's basically the idea, so it interpolates 84 00:09:22,400 --> 00:09:27,320 the values, and then it shows the interpolated image on the screen which 85 00:09:27,320 --> 00:09:37,220 looks pretty nice in our eyes. Okay, let's see this. And this is the image. 86 00:09:37,220 --> 00:09:43,790 And you see that this is quite stretched a little bit, so it was a tall image, 87 00:09:43,790 --> 00:09:50,810 but now it's quite wide, and because this is actually the width of the image 88 00:09:50,810 --> 00:09:58,910 and this is the height, so if you want you can see 500 and 1000. Save again. 89 00:09:58,910 --> 00:10:07,190 And now you see more or less the true ratio of the image, but if you want to keep the 90 00:10:07,190 --> 00:10:16,190 ratio of the image you'd want to go more advanced here, so let's say from this, 91 00:10:16,190 --> 00:10:21,680 so let's say we would want to show a size that is half of this so that that would 92 00:10:21,680 --> 00:10:26,450 keep the ratio of the image. So what we can do is we need to access to these 93 00:10:26,450 --> 00:10:32,870 values, and we can grab those values form the shape, the method, so this produces 94 00:10:32,870 --> 00:10:44,120 a tuple with these two values, and then we go here and say image shape, and this 95 00:10:44,120 --> 00:10:52,370 would be this value, so 990 which has an index of 1, and then we have again 96 00:10:52,370 --> 00:10:59,770 image dot shape with index of 0 for this number. 97 00:11:00,630 --> 00:11:09,100 And then we want to divide this by two. Okay, and I expect to get an error from 98 00:11:09,100 --> 00:11:15,820 this, but let's see, so yeah we've got an error. It's a type error. Integer argument 99 00:11:15,820 --> 00:11:23,980 expected got float, but here, what we're doing here is that when we divide this 100 00:11:23,980 --> 00:11:28,390 number by two we will get a float, so we would get something 101 00:11:28,390 --> 00:11:33,090 like 742.5 In that case what you want to do is 102 00:11:36,440 --> 00:11:49,550 convert this to an integer, so 742.5 becomes 742, okay and let's keep 103 00:11:49,550 --> 00:12:00,529 the consistency, so integers and for this as well. Let's see! We've got 104 00:12:00,529 --> 00:12:07,670 an invalid syntax here, so it points us to line 11 somewhere in the beginning which 105 00:12:07,670 --> 00:12:14,230 can be quite misleading, so you want to see before that line which is here and 106 00:12:14,230 --> 00:12:19,870 you can see that this bracket here close is here, so we need another bracket 107 00:12:19,870 --> 00:12:27,800 which closes in the first bracket here. So save that, try again and this time the 108 00:12:27,800 --> 00:12:34,940 image looks good. So you'll learn how to load an image in Python, and you'll learn 109 00:12:34,940 --> 00:12:40,160 how to resize an image, how to show an image on the screen and now let's go 110 00:12:40,160 --> 00:12:46,310 ahead and write the resized image in a new file. For that you'd want to use the 111 00:12:46,310 --> 00:12:52,850 imwrite method, so image write and you want to give a name to a new image. 112 00:12:52,850 --> 00:13:02,060 Galaxy let's say resized dot jpg and then you pass the image object that you 113 00:13:02,060 --> 00:13:09,080 want to store in this file. So the comma goes also here and the 114 00:13:09,080 --> 00:13:19,850 image you want the store is resized image. That's it. Execute and we got the 115 00:13:19,850 --> 00:13:28,100 old image resize on the fly, so Python gets a Numpy array, it interpolates it, 116 00:13:28,100 --> 00:13:32,930 so it resizes it and then it shows it on the screen, so this is galaxy this one here. 117 00:13:32,930 --> 00:13:42,490 And then we can close this and here we've got our new image, so galaxy resized. 118 00:13:43,400 --> 00:13:48,980 So if you go to the folder where these files are, you'll see that this image 119 00:13:48,980 --> 00:13:56,540 has new dimensions, so 495 by 742. And that's what I wanted to teach you 120 00:13:56,540 --> 00:14:00,000 in this lecture. Hope you enjoyed it and talk to you later.