Professional Documents
Culture Documents
Pic 8722
Pic 8722
Image Restoration
Spatial Domain
Freq. Domain
Source of Noise:
•Image acquisition (digitization) and/or transmission
•Image sensors and environmental conditions
Works well for salt and other noise but not good for pepper noise
Order-Statistics Filters
Median Filter Midpoint filter
Good for both bipolar and unipolar impulsive noise Good for random noise like Gaussian and Uniform
Max and Min filters Alpha-Trimmed mean filter
d=6
Adaptive Filters
Adaptive, local noise reduction filter
Adaptive Filters
Adaptive, local noise reduction filter
Blind Deconvolution
Observed subimage
The algorithm presented in this section has the notable feature that it
yields an optimal result for each image to which it is applied. Of course,
it is important to keep in mind that these optimality criteria, while
satisfying from a theoretical point of view, are not related to the
dynamics of visual perception. As a result, the choice of one algorithm
over the other will almost always be determined (at least partially) by the
perceived visual quality of the resulting image. By using the definition of
convolution, we can express in vector-matrix form, g = Hf + η. For
example, suppose that g(x,y) and f(x,y) are of size M×N. Then we can
form the first N elements of the vector g by using the image elements in
the first row of g(x, y), the next N elements from the second row and so
on. The resulting vector will have dimensions MN×1. These are also the
dimensions of f and η, as these vectors are formed in the same manner.
12/20/2020 Digital Image Processing - Ali Aghagolzadeh 62
Chapter 5
Image Restoration
The matrix H then has dimensions MN × MN. Its elements are given by
the elements of the convolution integral.
It would be reasonable to come to the conclusion that the restoration
problem can now be reduced to simple matrix manipulations.
Unfortunately, this is not the case. For instance, suppose that we are
working with images of medium size say M = N = 512. Then the vectors
given before would be of dimension 262,144 × 1, and matrix H would
be of dimensions 262,144 × 262,144. Manipulating vectors and matrices
of these sizes is not a trivial task. The problem is complicated further by
the fact H is highly sensitive to noise (after the experiences we had with
the effect of noise in the previous parts, this should not be a surprise).
However, formulating the restoration problem in matrix form does
facilitate derivation of restoration techniques.
12/20/2020 Digital Image Processing - Ali Aghagolzadeh 63
Chapter 5
Image Restoration
M 1 N 1
C 2 f ( x, y )
2
g Hfˆ
2
2
x 0 y 0
w wT w fˆ
2
M 1 N 1
C 2 f ( x, y )
2
x 0 y 0
2
g Hfˆ
2
a or
2 2
otherwise return to2 step 22 after increasing γ if r
decreasing γ if r a and use the new value of γ to
recompute the optimum estimate Fˆ (u , v ) .
---------------------------------------------------------------------------------------
Other procedure, such as a Newton-Raphson algorithm, can be used to improve the
speed of convergence. In order to use this algorithm, we need the quantities r 2 and
2
2
. To compute r , we note that . From which we
obtain r(x, y) by computing the inverse transform of R(u,v).
Computational of
2
Then leads to an
increasing result. First, consider the variance of the noise over the entire
image, which we estimate by the sample-average method:
where
2
is the sample mean. With reference to the form of r , we note that the
double summation in Eq. for is equal to . This gives us the
2 2
expression
The initial value of γ=10-5, the correction factor for adjusting γ was 10-6, and the value of α=0.25.The noise parameter
specified were the same used to generate Fig. 5.25. a noise variance of 10-5, and zero mean. The restored result is
almost as good as Fig. 28(c), which was obtained by Wiener filtering with K manually specified for the best visual
results. Fig. 5.31(b) shows what can happen if the wrong estimate of noise parameter are used. In this case, the noise
variance specified was 0.01 and the mean was left at a value of 0. The result in this case is considerably more blurred.
12/20/2020 Digital Image Processing - Ali Aghagolzadeh 73
Chapter 5
Image Restoration
with α and β being positive, real constant. The geometric mean filter
consists of the two expression in brackets raised to the power α and 1-α,
respectively. When α=1, this filter reduces to the inverse filter. With α=0,
the filter becomes the so-called Parametric Wiener Filter, which reduces
to the standard Wiener filter when β=1. If α =1/2, the filter becomes a
product of the two quantities raised to the same power, which is the
definition of the geometric mean, thus giving the filter its name.
12/20/2020 Digital Image Processing - Ali Aghagolzadeh 75
Chapter 5
Image Restoration
fˆ (0,0) , we substitute (x, y) = (0, 0) into the given Eqs. to obtain a pair of
coordinates (x’, y’) from those Eqs. to obtain a pair of coordinates (x’, y’) from those
equations. Then we let fˆ (0,0) g ( x ' , y ' ) , where x’ and y’ are the coordinate
values just obtained. Next, we substitute (x, y) = (0, 1) into the given Eqs., obtain
another pair of values (x’, y’), and let fˆ (0,1) g ( x ' , y ' ) for those coordinate
values. The procedure continues pixel by pixel and row by row until an array whose
size does not exceed the size of image g is obtained.
12/20/2020 Digital Image Processing - Ali Aghagolzadeh 80
Chapter 5
Image Restoration
Geometric Transformation - Spatial
A column (rather than a row) scan would yield identical results. Also,
a bookkeeping procedure is needed to keep track of which
quadrilateral apply at a given pixel location in order to use in order
to use the proper coefficients.
Tiepoints are established by a number of different techniques,
depending on the application. For instance, some image generation
systems having physical artifacts (such as metallic points) embedded
on the imaging sensor itself. These produce a known set of points
(called reseau marks) directly on the image as it is acquired. If the
image is distorted later by some other process (such as an image
display or image reconstruction process), then the image can be
geometrically corrected using the technique just described.
12/20/2020 Digital Image Processing - Ali Aghagolzadeh 81
Chapter 5
Image Restoration
Geometric Transformation – Gray-Level Interpolation