1
00:00:01,450 --> 00:00:04,280
So let's talk a bit about how images are formed.

2
00:00:04,390 --> 00:00:09,880
So we're going to look at the most basic example of an image formation onto a piece of film.

3
00:00:09,970 --> 00:00:15,520
So imagine you're outside in a park and you're holding a strip of film while facing a tree so light

4
00:00:15,520 --> 00:00:20,610
reflects off the tree at different points and bounces off a tree onto your piece of film.

5
00:00:21,100 --> 00:00:25,930
So as you can see in this example what happens here is at the top of the tree and the middle of the

6
00:00:25,930 --> 00:00:29,240
tree are going to reflect at similar points along the farm.

7
00:00:29,350 --> 00:00:30,350
That's not good.

8
00:00:30,480 --> 00:00:36,160
That will actually focus that will basically result in an unfocused image or blurred image here.

9
00:00:36,940 --> 00:00:41,150
So as we can see that's not how our eyes or cameras work.

10
00:00:42,430 --> 00:00:45,720
This is a best example of how our eyes and cameras work.

11
00:00:46,030 --> 00:00:53,140
We essentially use a barrier to block off most points of light while leaving a small gap here and that

12
00:00:53,140 --> 00:00:55,010
gap has called aperture.

13
00:00:55,300 --> 00:00:59,040
And this allows us some points of light to be reflected onto the form.

14
00:00:59,380 --> 00:01:05,460
So this gives you a much more focused image and that's actually the basis of a pinhole camera.

15
00:01:05,590 --> 00:01:11,190
So it is a problem with a simple pinhole camera model in that the aperture is always fixed.

16
00:01:11,230 --> 00:01:18,040
So that means that a constant amount of light is always entering this will which can be sometimes overpowering

17
00:01:18,130 --> 00:01:18,730
for the film.

18
00:01:18,730 --> 00:01:20,960
Meaning that everything is going to look white.

19
00:01:21,100 --> 00:01:28,210
And secondly we can focus using this fix up issue to focus the image even better although it's never

20
00:01:28,210 --> 00:01:30,190
going to be as bad as the previous image.

21
00:01:30,190 --> 00:01:32,700
We still need to move the film back and forth.

22
00:01:33,530 --> 00:01:35,780
And that's not really a good system.

23
00:01:35,780 --> 00:01:37,150
So how do we fix this.

24
00:01:37,250 --> 00:01:44,030
Well by using a lens and an adaptive lens which is what most modern cameras and our eyes use it allows

25
00:01:44,030 --> 00:01:49,700
us to control the aperture size and in photography aperture size is referred to as f stops and cameras

26
00:01:50,150 --> 00:01:57,650
and lowa is better and also allows us to get some nice depth of field which is also called booka in

27
00:01:57,650 --> 00:01:59,450
photography.

28
00:01:59,480 --> 00:02:03,940
Just so you know book Booker is a highly desirable trait in photography.

29
00:02:04,070 --> 00:02:09,110
It allows us to have very blurred backgrounds while we focus on a growing image resulting in a pretty

30
00:02:09,110 --> 00:02:11,130
nice effect.

31
00:02:11,150 --> 00:02:17,240
Secondly with using a lens you actually can control the lens wit which allows us to instead of moving

32
00:02:17,240 --> 00:02:18,460
the film back and forth.

33
00:02:18,770 --> 00:02:22,430
We actually use a lens to focus directly on this point here.

34
00:02:22,880 --> 00:02:28,490
This results in a very nice nicely controlled system so fiercely before discussing how computers two

35
00:02:28,490 --> 00:02:29,150
images.

36
00:02:29,150 --> 00:02:34,580
I think it's good to discuss how humans see images and it is one thing you shouldn't know humans are

37
00:02:34,580 --> 00:02:40,520
exceptionally good at Image Processing starting with our eyes they're remarkably good at focusing quickly

38
00:02:40,880 --> 00:02:47,330
seeing in varying light conditions and picking up sharp details and then in terms of to putting what

39
00:02:47,330 --> 00:02:54,440
we see humans are exceptional at this as we can quickly understand the context of different images and

40
00:02:54,440 --> 00:03:01,550
quickly identify objects faeces you name it we can actually do this far better than any computer vision

41
00:03:01,670 --> 00:03:02,930
technique right now.

42
00:03:03,940 --> 00:03:10,820
And our brain our brains do this by using six layers of visual processing that you can see here.

43
00:03:11,230 --> 00:03:15,430
I won't go into the details of this but it's incredibly complicated.

44
00:03:15,520 --> 00:03:20,160
And if you're curious you can visit the Wikipedia page on our visual system right here.