Chapter 9:

Question 1

What is a classification rule?

Answer 1

A rule that assigns pixels to a certain class. (c(v))

Question 2

Give example of training in pixel classification

Answer 2

Expert annotation and then histogram to see what grey level values are belonging to the different classes

Question 3

What is the minimum distance classifier?

Answer 3

Calculate a mean for each class - new pixels are assigned to the mean they are closest to.

Question 4

What is a parametric classification?

Answer 4

Trained classes are used to approximate a distribution that can tell something about the probability.

Question 5

How to calculate parametric classification intervals?

Answer 5

By calculating the intersect between the distributions (one distribution equal to the other distribution).

Question 6

Whatis bayesion classification?

Answer 6

If we knowsomething about the distribution of the different classes (we have priors) we can use this information to make classification better.

Question 7

Name the 5 steps in constructing a bayesian classifier:

Answer 7

1) identify number of classes
2) Mark chosen classes in training image
3) compute parametric description of each class (P(v|Ci))
4) Estimate prior probabilitis (P(ci)
5) Compute P(Ci | v) for every pixeland every class.