Neighbourhood Processing - Lecture 3 - Week 1

Question 1

What is rank filtering?

Answer 1

Take values under the filter window (kernel) and order them, then output based on some value, e.g. in median filtering (Maybe also rank filtering) the median value is used

Question 2

Why is the mean not as effective as the median in rank filtering?

Answer 2

Noise is lighter, but larger & edges are blurred

median removes noise and blurs edges less

Question 3

Why is convolution good?

Answer 3

Looks at spatial pattern of pixels, not just the value of the pixels

Question 4

In convolution what is the final sum divided by when smoothing?

Answer 4

The sum of the values in the kernel

Question 5

What padding methods exist for convolution?

Answer 5

Zero - set all pixels to zero

Constant (border colour) - set all pixels outside the source image to a specified border value

Mirror - reflect pixels across the image edge

Question 6

What does a high frequency mean in an imge?

Answer 6

abrupt changes in pixel value

Removing them smooths the image

Question 7

What are the parameters to gaussian blur?

Answer 7

Sigma (1/2 the size of the width of the distribution width)
K - the dimensions of the kernel (must be odd)