16-11-2012, 03:32 PM
Correspondence
Correspondence.pdf (Size: 538.51 KB / Downloads: 65)
INTRODUCTION
In image processing, images are often corrupted by positive and
negative impulses stemming from decoding errors or noisy channels.
Both are easily detected by the eye and degrade the image quality.
The nonlinear mean filter [2], [3] cannot remove such positive
and negative impulses simultaneously. The median filter performs
quite well, but it falters when the probability of impulse noise
occurrence becomes high. To overcome this situation, we propose
a new algorithm for adaptive median filters with variable window
size. This filter is to be robust in removing mixed impulses with high
probability of occurrence while preserving sharpness. This algorithm,
called the ranked-order based adaprit*e median jilter (RAMF), is
based on a two-level test. The first level tests for the presence of
residual impulses in the median filter output, and the second level
tests whether the center pixel itself is corrupted by an impulse or not.
In some image applications, it is frequently desirable to remove
noise that might be impulsive and/or nonimpulsive, with minimum
distortion of the original image information. One of the undesirable
properties of the median filter is that it does not provide sufficient
smoothing of nonimpulsive noise. To overcome this, various
techniques [I], [4] have been used.
SECONDN OISEM ODEL: SAMF
In the second model, the noise corrupted pixel is x , =~ s ,, +
where t i L J is iid impulsive noise having Laplacian or Cauchy or a
mixture of Gaussian and Cauchy distributions. This SAMF algorithm
in this instance detects the width of the impulse and adjusts the
window accordingly until the noise is eliminated.
CONCLUDINREGM ARKS
Based on two types of impulse noise corrupted image models, we
have introduced two new algorithms for adaptive median filter with
variable window size for removal of high density impulse noises
while preserving image sharpness: the ranked-order based adaptive
median filter (RAMF) and the impulse size based adaptive median
frlter (SAMF).
The RAMF, based on a two-level test, is simple in its operation and
removes positive and negative impulse noise simultaneously while
preserving sharpness better than the nonlinear mean L,, filter [2].
The SAMF, based on impulse noise size detection inside a window,
is simpler than Lin’s adaptive scheme [5] and removes high density
impulses, smoothes nonimpulsive noise, and preserves details better
than Lin’s scheme.
The simulation results also show that the performance of these
filters are superior to that of the median filter.