International Conference on Computer Science and Artificial Intelligence (ICCSAI 2013)
ISBN: 978-1-60595-132-4
The Study of Strong Salt-Pepper Noise Image Filter Processing Algorithm
Based on Center Minimum Distance
Bao-zhong Wang
School of Information Management, Shanxi University of
Finance & Economics
Taiyuan, China
Xuan-bing Qiu, Ji-lin Wei, Chao Wei
Department of Physics, Taiyuan University of Science &
Technology
Taiyuan, China
Corresponding Author: Xuanbing Qiu qiuxb@tyust.edu.cn
Gang Li
Testing laboratory,
Shan Xi Academy of Analytical Science
Taiyuan, China
Abstract—The image denoising is always a significant topic of
image preprocessing. The standard median filtering is an
efficient method to remove the Salt-Pepper (SP) noise from
image. However, the standard median filtering only
appropriate for weak noise situation, but doesn’t work to
strong SP noise. According to the feature of strong SP noise
image, this paper provides an image filtering algorithm which
based on center minimum distance. The algorithm removes the
strong SP noise efficiently and save the details of image
perfectly with the feature of easy implement, high self-
adaptability and high execute efficiency. The result of
experiment shows that the algorithm achieves the better
processing effect to SP noise image with the intensity range
from 10% to 95%.
Keywords- digital image; strong Salt and Pepper (SP) noise;
minimum distance; self-adaptability
CLC: TN911.73
I. INTRODUCTION
Noise is a kind of random variation of brightness or
color information in images [1]. Noise may appear in
engineering drawing images during data acquisition such as
scanning [2]. It leads to incorrect analysis and
comprehension of source information of received image
which captured from organ of vision or system sensor. The
image noise may generate from the bad sensor spot of image
capture device or the image transport of strong signal noise.
The SP noise does the serious disturbance in digital image
because the gray value of noise, which compared to original
image, fluctuates acutely. The noisy signal would directly
lead to unsharpness of total image, the fuzzy scenery and
background clutter. For example, in real application, the
rates of face image recognition will unavoidable decrease
because of noise disturbance. The noise always affects to
further processing of image. Therefore, one of the
significant subjects of image preprocessing is that how to
suppress noise better and save more image details. Median
filter is a well known method that can re-move salt/pepper
noise from images [3].
The traditional treatment of SP noise is the median
filtering based on rank-order filtering which proposed by
Tukey at the beginning of the 1970s [4]. However, the effect
of this method is limited by the size of filtering window.
Different remedies of the median filter have been proposed,
e.g., the adaptive median filter [5], the multistate median
filter [6], or the median filter based on homogeneity
information [7], [8]. The effect of denoising is not good
when the window is too small but the image become
blurring with details lost when window is too large. The
processing effect of standard median filtering will decrease
dramatically when it deals with the strong SP noise. People
raised a lot of schemes to reduce the strong SP noise of
image. The filtering algorithm of strong SP noise which
based on the theory of partial differential equations is
proposed by Chan R H in 2005, and it has gained better
effect [9]. Center Weighted Median (CWM) is a superior
en-hancement to Median filter [10]. Morphological
operators are another mean to pro-cess images using set
theory [11]. Another filter that can remove noise from
document images is the kFill filter [12] [13]. In 2007, the
execute time of algorithm is significantly shortened because
that Lei Zhu provides a kind of reliable and fast SP
removing algorithm with detail-preserving [14]. Bo Wang
presents a king of soft-threshold histogram weighted
filtering with correlativity for high density SP noise images,
using the feature of image neighborhood related and the
robustness of histogram to SP nois
e
[15]. In 2009, Jie Li
proposed an improved median filtering algorithm which
gained a better effect to strong SP noise image [16]. Gang
Li raised a kind of twice adaptive median filtering algorithm
with strong SP noise image and it received a better effect
[17]. Those algorithms all can gain the better effect to
remove the strong SP noise of image; but they are not suit to
embedded real-time system because of complex implement
processing, high complexity calculating and long calculating
time.
National Natural Science Foundation of China (61178067)