Lecture 2 Digital Image Analysis And Processing

Question 1

What are the three major types of image processing?

Answer 1

1. Local (each pixel will not interfere each other)
   - contrast and brightness adjustment
   - gamma correction
   - histogram equalisation
   - threshold
2. Neighbourhood
   - spatial filtering’
   - edge detection & enhancement
   - entropy filtering
3. Global
   - Fourier filtering

Question 2

What are the formulas for local processing (brightness adjustment / contrast adjustment / gamma correction)

Answer 2

Brightness adjustment:
Output = input + K
(Add a same constant)

Contrast adjustment
Output = input * m
(Multiply a same constant: I.e. the difference between the pixel value also doubled)

Gamma correction
Output = (input)^r

Question 3

What is a gray-scale histogram?

Answer 3

The histogram shows the distribution of pixel value

Question 4

What is bin width in gray-scale histogram?

Answer 4

The pixel values are classified into intervals to form the histogram. Bin width can be manipulated.

Question 5

What do window width and window level control?

Answer 5

The window allow the user to control the interval of gray levels to be displayed on the monitor in real time.
Level = centre of window
Width = range of window
Gray levels below min level : zero luminance
Gray levels above max of window: max luminance
Smaller window = increase display contrast

Question 6

What is intensity based segmentation (eg. Lung segmentation)

Answer 6

• algorithmically define the lower and upper limits of pixel intensity of each objects

Question 7

What are the limitation of intensity-based segmentation?

Answer 7

• assumed the intensity ranges of individual objects are disjoint (not many overlapping pixels)
• assumption of homogeneity (same structure have same CCT intensity value)

Question 8

What is the basic idea in correlation filtering?

Answer 8

1. Slide the centre of the correlation kern on the image
2. Multiply each weight in the correlation kernel by the pixel in the image
3. Sum these products and return the value to the central kernel

Question 9

What is the problem relating to the size of a filtered image?

Answer 9

Filtered images are always smaller than the original

Question 10

What method can solve the problem of reduction in size of correlation filtered images?

Answer 10

Assume boundary condition
(Make up boundaries number)
To make the filter image size = original image size

Question 11

Name the filter
1 2 1
0 0 0
-1 -2 -1

Answer 11

Edge detection filter
With edge width = 2 pixel

Question 12

Name the filter
1 1 1
1 1 1
1 1 1

Answer 12

Averaging filter
Result in blurry image

Question 13

Name the filter
1 4 7 4 1
4 16 26 16 4
7 26 41 26 7
4 16 26 16 4
1 4 7 4 1

Answer 13

Gaussian filter

Question 14

The domain of Gaussian function is __________

Answer 14

Infinitely large
But in practice it is nearly zero at more than about 3 D0 from the mean
Therefore the function can be reduced to a 5 x 5 matrix

Question 15

Name the filter
1 0. 0 1
0 -1. -1 0

Answer 15

Robert cross filtering

The direction of zero detect the edge direction
Diagonal edge detection

Question 16

Name the filter
1 2 1
0. 0. 0
-1 -2 -1

Answer 16

Sobel filtering which detects horizontal edges

Question 17

What is edge based segmentation?

Answer 17

It is an edge filtering technique that detect the border of a region and formed closed curves.

There is no assumption of homogeneity of the region, only the change in pixel values along the edges matters

Question 18

What are the limitation of edge-based segmentation?

Answer 18

• obtained edges are usually disconnected
• more sensitive to noise

Question 19

Name 1 non-linear spatial filter

Answer 19

Median filter