Hello everyone and welcome back to computer vision lecture series.

In this lecture we are going to learn about edge detection and how edge detection is important

for feature matching and feature extraction for higher level computer vision tasks.

So let's see.

In edge detection what we are going to cover is what type of edges exist, what kind of

influence does the noise have from the image over the edges and we are going to learn about

specific edge detector called cany edge detector.

Cany edge detector is very popular in computer vision literature and if you have taken computer

vision class and you don't know about cany edge detector then it's considered that you

don't really know that you have not really taken computer vision lectures.

So we are going to dig a bit deeper and go procedurally on into look procedurally into

how cany works, cany edge detector works.

When it was introduced it was the state of the art for edge detection techniques.

Before we go ahead for edge detection let's look at what do we mean by low level and high

level computer vision tasks.

So computer vision basically works in the domain of understanding real world images.

Through real world images we extract information that is relevant for our application.

It could be edges, it could be corners, it could be a particular feature and the tools

and techniques that you use they come under computer vision.

The computer vision pipeline has a whole set of steps from sensing to data representation,

from extraction of edges, corners up to 3D reconstructions and each of these steps have

different methods related to them and there are different techniques and state of the

art algorithms that you can use for accomplishing each and every of these steps.

These steps are not exhaustive however but as you go from sensing to 3D reconstruction

the definition of computer vision to be changes from lower level to higher level.

The reason being low level tasks are considered as the tasks which are very small and require

local feature understanding.

Whereas 3D reconstruction is considered high level computer vision tasks because you are

constructing a 3D perspective of your real world through different images.

And so starting from sensing until the data representation there are multitudes of steps

involved in between.

Before we start to look into how edges behave and what are the definition of edges we want

to learn about vanishing points and vanishing lines.

So what is a vanishing point basically?

Let's consider a ground plane.

A ground plane is any plane in the real world when you are taking an image of the real world.

This is your image plane and O is the camera center.

When you capture an image the lines which are parallel in the ground plane tend to meet

at a certain point in your image plane and this point is considered as vanishing point.

Any two parallel lines along this ground plane will have only one vanishing point.

That's the property of vanishing point.

Another property is the ray originating from the camera center going through this vanishing

point is parallel to these parallel lines lying on the ground plane.

As you can imagine if you change this ground plane you can get different vanishing points.

So there can be multiple vanishing points in your image plane.

Another aspect is, so this is an example of a vanishing point basically.

If we consider this as our ground plane there are these rail tracks are all these four rail

tracks are parallel lines and they meet at one point in the image.

That is the vanishing point.

The property of vanishing point is that it can be used to reconstruct camera or to get