Current time:0:00Total duration:7:09
0 energy points

Feature detection and parallel processing

In this video, I review our ability to break down an image into its component "features" such as color, form, and motion. This is known as feature detection, and since the detection of various features happens simultaneously, or in parallel, it is referred to as parallel processing.  By Ronald Sahyouni. Created by Ronald Sahyouni.
Video transcript
When you're looking at an object, it's necessary to break it down into its component features in order to make sense of what you're looking at. This is known as feature detection. There are many components that make up feature detection. So let's go into these. When you're looking at a rose, not only do you have to look and decide, OK, what color am I looking at, you also have to figure out, OK, what shape is the rose. So you have to get a little bit of information about the form of the rose. You also need to get information about motion. So is the rose moving? Am I throwing it across a room? What's going on? So whenever we look at any object, we need to get information about color, we need to get information about the form of the object, and information about motion. So let's go into each one of these different features. So our ability to sense color actually arises from the presence of cones within our retina. Cones are extremely important, because they're sensitive to various types of light. So we have three major types of cones. There are red comes, which make up 60% of the cones in your eye. There are green cones, which make up 30% of the cones in your eye. And there are blue cones, which make up about 10% of the cones in your eye. Now, we divide them into red, green, and blue cones because the red cones are extremely sensitive to red light. So if we're looking at this rose, the petals are actually reflecting red light. And so this is red light enters your eye. And if it happens to hit a red cone, then the cone will activate, and it will fire an action potential. And this action potential will reach your brain, and your brain will recognize that you're looking at something that is red. So this basically happens regardless of what we're looking at. And this has come to be known as the trichromatic theory of color vision. So what other features do we need to take into consideration when we're looking at this rose? So aside from breaking the rose down into the different colors, we also need to figure out, OK, what are the boundaries of the rose, so the boundaries of the stem, the boundaries of the leaf, the boundaries of the petals, from the background. And this is also really important, because not only do we need to distinguish the boundaries, but we also need to figure out, OK, what shape are the leaves, what shape are the petals. And these are all very important things that your brain ultimately is able to break down. So in order for us to figure out what the form of an object is, we use a very specialized pathway that exists in our brain, which is known as the parvo pathway. So the parvo pathway is responsible for figuring out what the shape of an object is. So another way to say this is that the parvo pathway is really good at spatial resolution. Let me write that down. So spatial resolution. And what spatial resolution means is that it's really good at figuring out what the boundaries of an object are, what the little details that make up the object are. So if something isn't moving, such as when you're looking at a picture or when you're looking at a rose, you're able to break down and look at the little tiny details. So you're able to look at the little veins that make up the rose leaf. You're able to see all the little nuances of the rose. And that's because you're using the parvo pathway, which has a really high level of spatial resolution. One negative aspect of the parvo pathway is that it has really poor temporal resolution. And what I mean by this is that temporal resolution is motion. So if a rose is in motion, if I throw it across the room, I can't really use the parvo pathway to track the rose. The parvo pathway is used for stationary objects to acquire high levels of detail of the object. But as soon as that object starts moving, you start losing your ability to identify tiny little details in the object. And you've probably noticed this. So when you're in a car, you're driving along, and there's a little Volkswagen Beetle driving right by you, it's going pretty slow. You can acquire a good number of details about the car, about the driver. But if you look down at the wheels, the wheels are spinning really, really quickly. And so if you try and look at the rims and figure out what design are the rims, it's really hard to figure that out. That's because the wheels are spinning so fast that it's really difficult to acquire any type of detail about the shape of the rims, about what they look like, and things like that. And finally, the parvo pathway also allows us to see things in color. So the parvo pathway not only allows us to acquire fine details about what we're looking at, but it also allows us to see in color. So if something is moving, we can't use the parvo pathway. But what we do use is the magno pathway. So we use the magno pathway in our brain. And the magno pathway is basically a set of specialized cells-- just like the parvo pathway-- that allow us to encode motion. So it allows to encode motion. And what that means is that it has really high temporal resolution. So as we've said, the parvo pathway has really low temporal resolution. The magno pathway has high temporal resolution, which means that if something is in motion we're able to track it, we're able to see an object moving. And in contrast to the parvo pathway, the magno pathway has very poor spatial resolution. So one way to think about this is that if I was looking at this rose using the parvo pathway, it's stationary, and I'm able to see all these fine details in the petals, all these fine details in the leaves. But if I was looking at the rose using the magno pathway, the rose would actually look a little bit like this. It would look very blurry. You would only be able to see this kind of blurry aspect of the rose. And when you're looking at the stem and the petals, you only see this very blurry vision of the rose. This is one way to imagine the difference between parvo and magno. If you're looking at this rose using your parvo pathway, you see all these details. If you're looking at the same rose using the magno pathway, it's very blurry. But the benefit of the magno pathway is that if this rose were to be moving, you can see it move. Whereas if you're using the parvo pathway and a rose is moving, you wouldn't even be able to see it at all. These are kind of the pros and cons of the two pathways. And finally, the magno pathway does not encode any color. So it just simply would encode its motion. So our ability to detect these three different features whenever we're looking at an object all happens at the same exact time. So when I'm looking at a rose, I don't first focus on the color and then the form and then whether or not it's moving. I get all this information at the same exact time. And our ability to see all three things whenever we're looking at something at the same exact time is known as parallel processing.