Derivatives and Edges

Question 1

What are features in images typically characterised by?

Answer 1

A change in intensity

Question 2

Can you calculate the 1st derivative by convolution?

Answer 2

Not convolution alone.

Question 3

What two derivative filters are most commonly associated with edge detection?

Answer 3

-Roberts’ Cross Operators
-Sobel Operators

Question 4

What does unsharp masking do?

Answer 4

Make edges more enhanced without affecting the rest of the image much

Question 5

What are the 3 steps of unsharp masking?

Answer 5

• gaussian smoothing the original image
• subtracting smoothed version from original to get the unsharp mask
• adding the mask to make edges more obvious

Question 6

What is a downsize of unsharp filtering?

Answer 6

The effect is hard to predict so its hard to parameterise

Question 7

What method of edge highlighting is more direct than unsharp filtering?

Answer 7

Estimating derivatives

Question 8

Why is the second derivative more useful than the 1st in image processing?

Answer 8

• Stronger response to fine detail and simpler implementation in 2D

Question 9

What is a common (derivative) sharpening filter?

Answer 9

The Laplacian

Question 10

Simply - what is edge detection?

Answer 10

The process of marking the points at which image intensity changes sharply

Question 11

If an edge is a sharp rise in an intensity graph, what happens on a graph of the 1st derivative? on a graph of the second derivative?

Answer 11

1st - a peak
2nd - a zero crossing

Question 12

How is thresholding used in 1st derivative edge detection?

Answer 12

Significant peaks in values of the 1st derivative are high, so a threshold value is picked and all 1st derivatives above that are taken as the edge

Question 13

The 1st derivative is normally used for….. while the second is normally used for..?

Answer 13

1st - edge detection
2nd - sharpening