An Improved Median Filter Algorithm B
ased on VC in Image Denoising
Shulei Wu
1,3
,Xiangxiang Xu
1,2
, Haixia Long
1
1
College of Information Science and Technology
Hainan Normal University, HaiKou, China
2
Guizhou Qianchi Information services ltd,
GuiYang, China
e-mail: wushulei@netease.com
Huandong Chen
1,*
, Wenjuan Jiang
1
, Dong Xu
1
3
College of Information Science and Technology
Hainan University, HaiKou, China
* Corresponding author e-mail:
ch_huandong@163.com
Abstract—In this paper, we propose an improved median
filtering algorithm by adding filtration function before
replacing the value of the median position with the median and
doing multiple processing of median filtering to overcome the
shortcoming of the traditional Median Filtering Algorithm.
Experimental results are shown that an image after processed
by the improved algorithm is hard to find image noise, and has
good effect on saving the detail as well. It turns out that the
improved algorithm is better than the traditional Median
Filtering Algorithm.
Keywords-filtration funtion; median filter algorithm;median
filtering; image denoising
I. INTRODUCTION
There are many types of common image noise [1], such
as Gaussian pulse (salt and pepper), Rayleigh, gamma, etc.
Image noise is caused during transmission or digitization. To
reduce image noise and recover a clear image, it needs to use
denoising filter to process the image data.
Median filtering algorithm [2], proposed by Tukey in
1971, is one of the typical nonlinear algorithms based on
spatial domain to remove salt and pepper noise. Justusson [3]
proposed center weighted median filtering algorithm and
weighted median filtering algorithm. Although there are
many researches [4]-[7] working on denoising in spatial
domain, it is ineffective in preserving the detail of the image
and cannot get the desired result.
In this paper, we propose an improved median filtering
algorithm by adding filtration function before replacing the
value of the median position with the median and multiple
processing image data using median filtering to overcome
the shortcoming of the traditional Median Filtering
Algorithm. The remainder of this paper is organized as
follows. First, we discuss the median filtering algorithm in
Section Ⅱ. Next, we introduce an improved median filtering
method in Section Ⅲ. The results of our experiments and the
analysis are described in section Ⅳ . Finally, Section Ⅴ
summarizes this paper and makes some closing remarks.
II.
MEDIAN FILTER ALGORITHM
Median filter [2] is a nonlinear method of image
processing method, which has a good effect on filtering out
the noise. The basic principle of median filter algorithm is
that the value of a point in the digital image is replaced with
the median value of all the points in its neighborhood. Its
implementation is to get a matrix 3x3 or 5x5, then sort these
data in a vector and find out intermediate value, next put the
median value into the interm ediate position of this 3x3 or
5x5 matrix, then move on to the next position and loop this
process. The experiments later turn out that the size of the
filter matrix will have a certain influence on the result of
filtering noise. If filter matrix is too big, which will make the
edges blurred. Conversel y, if filter matrix is too small, which
is not good for de-noising. The process of median filtering is
shown in Figure 1.
Figure 1. Median filter method using 3x3 matrix.
For multi-color image, the first step is to separate the
image into multiple channels pre-processing. The second
step is to filter out noises for each channel. The last step is to
merge the image channels after processing into a complete
color image. In this paper, we use 3×3and5×5 matrices to
process image data and compare the results between them.
III. I
MPROVED MEDIAN FILTER ALGORITHM
It is not good to use traditional median filtering algorithm
to process image as shown in Figure 3. From Figure 3 (b),
we can observe that most of the noises can be filtered out.
But there is a small part of noised that has not been filtered.
So we m ake multiple processing image data using median
filtering. However, it is a poor effect in preserving image
details. Also, because the noise and the edge points are the
pixel gray level change in sharp, when median filter
algorithm changes the gray value of noise point, edge pixels
are changed in a certain extent at the same time. But the
value of noise point is almost extreme value in the pixels of
2014 10th International Conference on Computational Intelligence and Security
978-1-4799-7434-4/14 $31.00 © 2014 IEEE
DOI 10.1109/.90
193
2014 10th International Conference on Computational Intelligence and Security
978-1-4799-7434-4/14 $31.00 © 2014 IEEE
DOI 10.1109/CIS.2014.91
193