Download presentation
Presentation is loading. Please wait.
Published byHugh Hampton Modified over 9 years ago
1
1 © 2010 Cengage Learning Engineering. All Rights Reserved. 1 Introduction to Digital Image Processing with MATLAB ® Asia Edition McAndrew ‧ Wang ‧ Tseng Chapter 8: Image Restoration
2
2 © 2010 Cengage Learning Engineering. All Rights Reserved. 8.1 Introduction Image restoration concerns the removal or reduction of degradations that have occurred during the acquisition of the image Some restoration techniques can be performed very successfully using neighborhood operations, while others require the use of frequency domain processes Ch8-p.191
3
3 8.1.1 A Model of Image Degradation © 2010 Cengage Learning Engineering. All Rights Reserved. f(x, y) : image h(x, y) : spatial filter Where the symbol * represents convolution In practice, the noise n(x, y) must be considered Ch8-p.191
4
4 8.1.1 A Model of Image Degradation © 2010 Cengage Learning Engineering. All Rights Reserved. We can perform the same operations in the frequency domain, where convolution is replaced by multiplication If we knew the values of H and N, we could recover F by writing the above equation as this approach may not be practical Ch8-p.192
5
5 8.2 Noise © 2010 Cengage Learning Engineering. All Rights Reserved. Noise—any degradation in the image signal caused by external disturbance These errors will appear on the image output in different ways depending on the type of disturbance in the signal Usually we know what type of errors to expect and the type of noise on the image; hence, we can choose the most appropriate method for reducing the effects Ch8-p.192
6
6 8.2.1 Salt and Pepper Noise © 2010 Cengage Learning Engineering. All Rights Reserved. Also called impulse noise, shot noise, or binary noise, salt and pepper degradation can be caused by sharp, sudden disturbances in the image signal Its appearance is randomly scattered white or black (or both) pixels over the image Ch8-p.192
7
7 FIGURE 8.1 © 2010 Cengage Learning Engineering. All Rights Reserved. >> imshow(t)>> figure, imshow(t_sp) Ch8-p.193
8
8 8.2.2 Gaussian Noise © 2010 Cengage Learning Engineering. All Rights Reserved. Gaussian noise is an idealized form of white noise, which is caused by random fluctuations in the signal If the image is represented as I, and the Gaussian noise by N, then we can model a noisy image by simply adding the two Ch8-p.193
9
9 8.2.3 Speckle Noise © 2010 Cengage Learning Engineering. All Rights Reserved. Speckle noise (or more simply just speckle) can be modeled by random values multiplied by pixel values It is also called multiplicative noise imnoise can produce speckle Ch8-p.194
10
10 FIGURE 8.2 © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.194
11
11 8.2.4 Periodic Noise © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.195
12
12 8.3 Cleaning Salt and Pepper Noise © 2010 Cengage Learning Engineering. All Rights Reserved. Low-Pass Filtering Ch8-p.196
13
13 8.3.2 Median Filtering © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.197
14
14 FIGURE 8.6 © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.198
15
15 FIGURE 8.7 © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.198
16
16 8.3.3 Rank-Order Filtering © 2010 Cengage Learning Engineering. All Rights Reserved. Median filtering is a special case of a more general process called rank-order filtering A mask as 3×3 cross shape Ch8-p.199 0 1 0 1 1 1 0 1 0
17
17 8.3.4 An Outlier Method © 2010 Cengage Learning Engineering. All Rights Reserved. Applying the median filter can in general be a slow operation: each pixel requires the sorting of at least nine values Outlier Method Choose a threshold value D For a given pixel, compare its value p with the mean m of the values of its eight neighbors If |p − m| > D, then classify the pixel as noisy, otherwise not If the pixel is noisy, replace its value with m ; otherwise leave its value unchanged Ch8-p.199
18
18 FIGURE 8.8 © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.200
19
19 FIGURE 8.9 © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.201
20
20 8.4 Cleaning Gaussian Noise © 2010 Cengage Learning Engineering. All Rights Reserved. Image Averaging suppose we have 100 copies of our image, each with noise Because N i is normally distributed with mean 0, it can be readily shown that the mean of all the N i ’s will be close to zero The greater the number of N i ’s; the closer to zero Ch8-p.202
21
21 FIGURE 8.10 © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.203
22
22 8.4.2 Average Filtering © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.203
23
23 8.4.3 Adaptive Filtering © 2010 Cengage Learning Engineering. All Rights Reserved. Adaptive filters are a class of filters that change their characteristics according to the values of the grayscales under the mask Minimum mean-square error filter The noise may not be normally distributed with mean 0 Ch8-p.204
24
24 8.4.3 Adaptive Filtering © 2010 Cengage Learning Engineering. All Rights Reserved. Wiener filters ( wiener2 ) Ch8-p.205
25
25 FIGURE 8.12 © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.206
26
26 FIGURE 8.13 © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.206
27
27 8.5 Removal of Periodic Noise © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.207
28
28 FIGURE 8.15 © 2010 Cengage Learning Engineering. All Rights Reserved. BAND REJECT FILTERING Ch8-p.208
29
29 FIGURE 8.16 © 2010 Cengage Learning Engineering. All Rights Reserved. NOTCH FILTERING Ch8-p.209
30
30 8.6 Inverse Filtering © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.210
31
31 FIGURE 8.18 © 2010 Cengage Learning Engineering. All Rights Reserved. cutoff radius: 60 Ch8-p.211
32
32 FIGURE 8.18 © 2010 Cengage Learning Engineering. All Rights Reserved. cutoff radius: 80cutoff radius: 100 Ch8-p.212
33
33 FIGURE 8.19 © 2010 Cengage Learning Engineering. All Rights Reserved. d = 0.005 Ch8-p.213
34
34 FIGURE 8.19 © 2010 Cengage Learning Engineering. All Rights Reserved. d = 0.002 d = 0.001 Ch8-p.213
35
35 8.6.1 Motion Deblurring © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.214
36
36 8.6.1 Motion Deblurring © 2010 Cengage Learning Engineering. All Rights Reserved. To deblur the image, we need to divide its transform by the transform corresponding to the blur filter This means that we first must create a matrix corresponding to the transform of the blur Ch8-p.215
37
37 FIGURE 8.21 © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.215
38
38 8.7 Wiener Filtering © 2010 Cengage Learning Engineering. All Rights Reserved. Ch8-p.216
39
39 FIGURE 8.22 © 2010 Cengage Learning Engineering. All Rights Reserved. K = 0.001 Ch8-p.217
40
40 FIGURE 8.22 © 2010 Cengage Learning Engineering. All Rights Reserved. K = 0.0001K = 0.00001 Ch8-p.217
Similar presentations
© 2024 SlidePlayer.com Inc.
All rights reserved.