Main Examination Period 2021 — May — Semester B
ECS776P IMAGE PROCESSING
Duration: 3 hours
Note: Answer all the FOUR questions using font size no smaller than 11 and the answer to each question must be done in no more than 4 pages.
Question 1 (Point Processing)
Consider the 4x5 image with a dynamic range of [0-9] as in Figure 1:
| 
				 S1  | 
			
				 S5  | 
			
				 S9  | 
			
				 S4  | 
			
				 S2  | 
		
| 
				 S2  | 
			
				 S6  | 
			
				 S1  | 
			
				 S5  | 
			
				 S1  | 
		
| 
				 S3  | 
			
				 S7  | 
			
				 S2  | 
			
				 S6  | 
			
				 S8  | 
		
| 
				 S4  | 
			
				 S8  | 
			
				 S3  | 
			
				 S1  | 
			
				 S9  | 
		
Figure 1: Image of size 4x5
where S1-S2-S3- …S8-S9 are the 9 digits of your student number, from left to right.
For example, if your student number is 123456789, the original image you use for this question is
| 
				 1  | 
			
				 5  | 
			
				 9  | 
			
				 4  | 
			
				 2  | 
		
| 
				 2  | 
			
				 6  | 
			
				 1  | 
			
				 5  | 
			
				 1  | 
		
| 
				 3  | 
			
				 7  | 
			
				 2  | 
			
				 6  | 
			
				 8  | 
		
| 
				 4  | 
			
				 8  | 
			
				 3  | 
			
				 1  | 
			
				 9  | 
		
(a) Find the transformation for histogram equalisation of the original image and give the resulting image after histogram equalisation has been applied. [10 Marks]
(b) By using the histogram of the image obtained in (a), describe a simple method to threshold the image and give the result. [5 Marks]
(c) Find the mean and the standard deviation of the image pixel values by using the histogram of the image obtained in (a). [6 Marks]
(d) For an increasing function of point transformation, s = af(r + b), where r and s are the input and output intensities, respectively, and a and b are two constants. If we wish to preserve the intensity dynamic range [0,255], how can we find the two constants a and b? What are the values for a and b iff(r) = log10 (r)? [4 Marks]
Question 2 (Image Coding)
Consider the 4x5 image with a dynamic range of [0-9] as in Figure 2:
| 
				 S1  | 
			
				 S5  | 
			
				 S9  | 
			
				 S4  | 
			
				 S2  | 
		
| 
				 S2  | 
			
				 S6  | 
			
				 S1  | 
			
				 S5  | 
			
				 S1  | 
		
| 
				 S3  | 
			
				 S7  | 
			
				 S2  | 
			
				 S6  | 
			
				 S8  | 
		
| 
				 S4  | 
			
				 S8  | 
			
				 S3  | 
			
				 S1  | 
			
				 S9  | 
		
Figure 2: Image of size 4x5
where S1-S2-S3- …S8-S9 are the 9 digits of your student number, from left to right.
For example, if your student number is 123456789, the original image you use for this question is
| 
				 1  | 
			
				 5  | 
			
				 9  | 
			
				 4  | 
			
				 2  | 
		
| 
				 2  | 
			
				 6  | 
			
				 1  | 
			
				 5  | 
			
				 1  | 
		
| 
				 3  | 
			
				 7  | 
			
				 2  | 
			
				 6  | 
			
				 8  | 
		
| 
				 4  | 
			
				 8  | 
			
				 3  | 
			
				 1  | 
			
				 9  | 
		
(a) Give the mathematical representation of Gray code for natural binary number
bn−1bn−2 ⋯ b3 b2 b1 b0 , and find the bit-planes of the Gray code of the original image in Figure 2. [9 Marks]
(b) Calculate the entropy of the image in Figure 2. [5 Marks]
(c) Find the Huffman code of the image in Figure 2. [11 Marks]
Question 3 (General image processing operations and image sampling)
(a) In Figure 3, the first image (0) is the original image and all the other images (1)-(7) are the results of the application of some image processing operations on (0). Identify what operation has been used to produce each of the images (1)-(7) from image (0), and explain your answers.
Figure 3: Original image (0) and the processed results (1-7) [15 Marks]
(b) Give two methods in mathematical terms to transform. pixel values after image enhancement to a dynamic range from 0 to 255 for display. Explain the difference between the two methods. [5 Marks]
(c) Suppose that a satellite image has the pixel size of 10 metres by 10 metres. Answer the following questions:
i) What is the size of the image for a ground area of 12300 metres by 23400 metres?
ii) If the image has 2 components, the dynamic range for one component is from 100 to 200 , and the dynamic range for the other component is from 200 to 400, how many megabytes do we need to store the whole image without compression? [5 Marks]
Question 4 (image filtering)
(a) Show how a 3x3 mask can be obtained to approximate the second-order derivative of an
image f(x, y), [5 Marks]
(b) For a piece of a line in an image, [3,2,4,3,13,12,14,13], show the result of the convolution with the one-dimensional version of the mask obtained in (a). Indicate where the edges would be detected and explain why. Use replicate border extension for the convolution. [8 Marks]
(c) What is the Laplacian of Gaussian? What is the Difference of Gaussians? What is the difference between LoG and DoG filters? [6 Marks]
(d) Use your student number to fill in the following 3x3 image below
| 
				 S1  | 
			
				 S2  | 
			
				 S3  | 
		
| 
				 S4  | 
			
				 S5  | 
			
				 S6  | 
		
| 
				 S7  | 
			
				 S8  | 
			
				 S9  | 
		
Figure 4: Image of size 3x3
where S1-S2-S3- …S8-S9 are the 9 digits of your student number, from left to right.
For example, if your student number is 123456789, the original image you use for this question is
| 
				 1  | 
			
				 2  | 
			
				 3  | 
		
| 
				 4  | 
			
				 5  | 
			
				 6  | 
		
| 
				 7  | 
			
				 8  | 
			
				 9  | 
		
Then give the results for 2-pixel border extension with the image by using three border extension methods. [6 Marks]
	
 
	
 
	
 
	
版权所有:编程辅导网 2021 All Rights Reserved 联系方式:QQ:99515681 微信:codinghelp  电子信箱:99515681@qq.com  
免责声明:本站部分内容从网络整理而来,只供参考!如有版权问题可联系本站删除。