Review Problems for Midterm Examination: CS474/674
1. Draw a piecewise linear transformation that takes [0,39] to 0,
[40,200] to [0,255] and [201, 255] to 255.
2. How can the contrast of an image be increased for the middle range
of gray levels by using XView or LView Pro or Matlab? Give complete
directions.
3. If we wanted high compression to yield a much smaller image file,
and the image has continuous shades rather than sharp lines, what
format should we use to save an image? What are its disadvantages?
4. If we wanted to compress an image but wanted to keep the same level of
detail and not lose any information, what formats could we use for
saving the image?
5. What determines the resolution of an image when a very high resolution
image is scanned into digital form with a scanner (that uses diodes
to capture the light intensity in a small area)? What does the value
of each pixel value represent in terms of the original image?
6. If we use 8-bits for each pixel value for a color image, then what
do the pixel values represent in terms of colors (the R, G and B
values for the color guns for displaying the image on a video
monitor)?
7. Why is it so convenient to process gray scale images in the PGM format?
8. What does a histogram of a gray scale image tell us about an image that
has a bad distribution of gray levels? Explain.
9. What is the purpose of histogram equalization? Explain.
10. Suppose that you want to use LView Pro to filter an image with a
certain convolution mask. Describe how to do it, once the image is
loaded into LView Pro's workspace.
11. What is the JPEG image format good for? Explain.
12. Derive a mask for unsharp masking by adding two masks and give the
purpose of each of the two masks.
13. Design a 5x5 mask that detects edge pixels going in a diagonal
direction from the lower left to the upper right.
14. Design a 5x5 mask that will detect edge or line pixels moving in
either direction along a line from the upper right to the lower
left corners.
15. Draw a 3x3 mask and a 3x3 nbhd of a pixel. Give the mathematical
equation for the mask convolution of these two 3x3 blocks.
16. Design an algorithm that smooths pixels that have very different
values from the average value of the other nbhd pixels, but
otherwise does not change the pixel value.
17. Design a blurring mask that blurs moderately but not strongly.
18. Suppose we want to use the Laplacian convolution mask on an image
(all -1 values except for the value 8 in the center). What could we
do first to help keep it from accentuating the noise?
19. Describe how to do unsharp masking on the image "shuttle.pgm" by
use of either the tool LView Pro or Matlab.
20. If we load a PGM image into a text editor, how can we tell if it
is in packed format or not?
21. Draw a thresholding graph for a transformation that will segment
an image into regions of 5 separate gray scales.
22. Describe an algorithm in psuedo-code that will do the thresholding
of an image into 3 separate gray levels (describe only the critical
part where the thresholding is done).
23. Describe an algorithm that selects a pixel at random, and then
grows a connected region around it consisting of all pixel values
that are close to its value.
24. In Problem 23 above, what would cause the region to quit growing?
25. Write the pseudo-code for an algorithm that examines each pixel
in an image and computes the 8 differences of between this center
pixel and the other 8 pixels in its 3x3 nbhd. If the average
difference magnitude is greater than 12, then change output
center pixel value to have a greater difference (edge enhancement),
else put the output center pixel value to be equal to the average
value.
26. Does the process in Problem 25 above smooth or sharpen an image,
or both?
27. What is the net result if an image is converted to its negative
and then added to the original image (addition is on each pixel)?
28. What is the net result if an image is added to itself and truncated
so that the output sum pixel values are truncated to [0, 255]?
29. Give examples of 3x3 convolution masks that smooth and sharpen.
30. Give examples of 3x3 convolution masks that detect edges moving
along horizontal and vertical lines, moving left or right, or
moving up or down (both directions on each line). Give an example
of a mask that will detect edges in all directions (there are 8
directions, total).
31. Construct a 5x5 mask that detects edges across the line from
the upper left corner to the lower right corner. It must detect
edges in both directions. Do not use the center value, that is,
put 0 in the center position. Explain how it works.
32. Develop a processing function whereby a pixel is smoothed if all
pixels in its neighborhood are close to its gray level but is
sharpened if a majority of the neighborhood pixels differ
significantly. This smooths and sharpens at the same time.
33. Discuss the effect of unsharp masking when ß > 1 and the
smoothing mask to be subtracted is multiplied by a
= ß - 1.
34. What is the result of adding a sharpening and a smoothing mask to
obtain a resulting mask such that the entries of the resulting mask
sum to 1?
35. Develop a 5x5 Laplacian type mask by adding line detector masks for the
horizontal, vertical, and diagonal directions.
36. Complete the mask equation given just before Figure 5b.6 in Unit 5b
and explain the effects of each of the two parts.
37. Run XV or Lview Pro on shuttle.pgm and invert it by changing
the diagonal line in the Intensity box in XView
Windows | Color Edit
window, or by using
Color | Curves
in LView Pro).
38. Use XV or LView Pro to threshold the inverted image of Problem 37 to
convert it to black and white.
39. Write a simple program in C that performs mask convolution.
40. Perform unsharp masking on "Lena" to sharpen it by entering a single
mask into the program of Problem 39 above. What effect does ß
have (try a higher and lower value than ß = 2)? What effect does
the degree of smoothing have (for example, if the subtracted mask blurs
the image)?