SVIATOSLAV
VOLOSHYNOVSKIY
STOCHASTIC
IMAGE PROCESSING
|
|
STOCHASTIC
IMAGE RESTORATION
   
    
|
Defocusing
Degradation model:
y = Hx + n,
where y -
blurred image, x - original image, n
- additive noise, H - distortion operator.
Additive white Gaussian noise
 |
 |
 |
 |
|
Original image
|
Defocused image
(radius = 9)
|
Adaptive Tikhonov
regularization
|
Developed
Penalized ML
|
|
 |
 |
 |
| Original image |
Defocused image
(radius = 50) |
Adaptive Tikhonov
regularization |
Developed
Penalized ML |
Additive mixture noise (white Gaussian and
Laplacian noise)
|
 |
 |
 |
| Original image |
Defocused image (radius = 9)
and corrupted by Gaussian
and Laplacian (5%) noise
|
Robust adaptive
Tikhonov regularization
|
Developed
Penalized ML
|
|
 |
 |
 |
| Original image |
Defocused image (radius = 9)
and corrupted by Gaussian
and Laplacian (50%) noise
|
Robust adaptive
Tikhonov regularization
|
Developed
Penalized ML |
References
:
1.
S. Voloshynovskiy,
Iterative image restoration with adaptive regularization and parametric
constraints,
Journal of Image Processing & Communications,
3, 3-4, pp. 73-88, 1997.
|
|
|
If you have any questions
or suggestions, please send e-mail: svolos@cui.unige.ch
Copyright © 1996 - 2001 |
