CS 659讲解、辅导Image Processing、MATLAB程序讲解、辅导MATLAB设计

2019.02.16 - 首页 >> Matlab编程

Dr. Frank Shih

CS 659 Image Processing

Homework #2

Covering Lectures 4, 5, 6

NOTES: Submit only the homework “solution” (do not include these homework questions) in

Microsoft Word format to http://moodle.njit.edu/ before the above deadline. Absolutely, no late

submission is accepted. First, write your NAME and ID#. Write the answers in your own words

individually. Any plagiarism will post a “ZERO” score or cause a “FAIL” grade.

Submit your answer in a Microsoft Word file. Do not include the questions; just provide the

required MATLAB codes and answers in the file.

Totally, there are 5 questions. Each question is 20 points. Grading policy checks the correctness

and completion of showing resulting images, Matlab codes, and text responses.

Required reading material: HW2R.

Required images: barbara_gray.jpg, lena_256.bmp, text.jpg, njit_logo.jpg, Lin.jpg

NAME:

ID #:

2.1 (20 points)

(a) Refer to Example 2.6 of the reading material HW2R. Hand calculates the convolution and

correlation of the two images A by B shown below. Show step-by-step calculations and both

results. Treat the region outside of A as zero. Use the center pixel of B as the origin. Note: the

resulting convolution and correlation should be the same size as A. Do not use MATLAB for

this.

Image A:

0 1 0 0

1 2 1 3

1 1 0 1

0 1 1 0

Image B:

0 3 0

2 1 0

1 0 1

(b) Refer to Example 3.3 of the reading material HW2R. An image with 8 gray levels is shown in

the following figure. Hand calculate to apply histogram equalization to the image; draw the

histograms before and after equalization, and display the resulting data matrix. Do not use

MATLAB program. Show your step-by-step calculation procedure.

44444

3 4 5 4 3

3 5 5 5 3

3 4 5 4 3

44444

will look too dark. Then use histogram equalization to the dark image. An example of Matlab

code: x1=imdivide(x,3); x2=histeq(x1); imhist(x1); imhist(x2); Provide your source code, and

Dr. Frank Shih

show four images using subplot into a 2x2 window with title on each. First, the dark image;

second, the resulting equalized image; third, the histogram of the dark image; fourth, the

histogram after equalization.

2.2 (20 points)

Show Matlab code and resulting images.

(a) Read the image: lena_256.bmp. Corrupt the image using ‘salt-and-pepper’ noise with parameter

0.7 (MATLAB code: imnoise). Now apply a 5 x 5 filter of ones (divided by 25), a 7 x 7 filter of

ones (divided by 49), and a median filter (MATLAB code: medfilt2) of 5 x 5 and 7 x 7 to the

corrupted image. Use subplot to show 6 images in a 3x2 window. Show the original image, the

salt-and-pepper noise image, and four resulting images.

(b) Show two output images. Read the image: lena_256.bmp. First, corrupt the original image using

Gaussian noise with mean 0 and variance 1. Second, corrupt the original image using

salt-and-pepper noise with parameter 0.5.

“motion” blur representing motion in a 11-pixel neighborhood and at 60°. Second, show the

original image is blurred by “out-of-focus” blur with the circle of confusion radius of 11 pixels.

2.3 (20 points)

(a) Let f and g be two matrices as given below. Hand calculates the resulting matrix of using image

matching equation:

f g f g * '* . What is the output pixel value of the exact matching location?

You need show the step-by-step calculation process.

f =

0 1 0

1 1 1

0 1 0

g =

0 0 0 0 0 0 0 0

0 0 1 0 0 0 0 0

0 1 1 1 0 0 0 0

0 1 1 1 1 0 0 0

0 0 1 1 0 0 0 0

0 0 0 0 0 1 0 0

0 0 0 0 1 1 1 0

0 0 0 0 0 1 0 0

(b) Read the image “text.jpg”. Design a template to extract the character “X”. Hint: You can crop

the character image “X”, and perform correlation (imfilter) or the above method for pattern

matching. Or you can design your own method. Show your method, Matlab code, resulting images,

and the row-column position of exacting matching.

Dr. Frank Shih

2.4 (20 points), Image Slicing

(a) Use the image: lena_256.bmp. Perform 4-level quantization. It means the original gray value 0

to 63 is assigned to an output value 0, 64 to 127 is assigned to an output value 64, 128 to 191 is

assigned to an output value 128, and 192 to 255 is assigned to an output value 192. Implement the

4-level quantization and display the quantized image.

(b) Use the image: lena_256.bmp. Perform bit plane slicing. Since it is a 8 bit gray-scale image,

generate and display 8 bit plane images in a subplot. Which bit plane is the most significant one that

describes the general image content?

2.5 (20 points), Image Subtraction

(a) Background subtraction is used to analyze the foreground regions. Show four images in a

subplot. First, “back1.jpg” is a whiteboard image with a mathematical equation written and taken in

very poor illumination conditions. Second, “back2.jpg” is another snapshot of the same whiteboard

without any text on it. Third, compute the absolute difference of these two images. Fourth, use the

complementary image of the previous result to visually reveal the written equation on the

whiteboard. Helper functions: imabsdiff, imcomplement.

(b) Subtractions are also useful to locate the change of blindness. Show four images in a subplot. (1)

and (2), “chng1.jpg” and “chng2.jpg” are two scenes that differ in another time instance. (3)

compute the absolute difference of these two images. (4) use the complementary image of the

previous result to visually reveal the changes even more prominently.

lena_256.bmp. Generate and display images in a subplot: (1) The original image. (2) Subtract a

Gaussian blurred image from the image itself with a kernel of 3 x 3 and variance of 1. (3) Subtract a

Gaussian blurred image from the image itself with a kernel of 10 x 10 and variance of 9. (4)

Subtract the 10 x 10 blurred image from the 3 x 3 blurred image.