STUDY ON METHODS OF NOISE REDUCTION IN A STRIPPED IMAGE 
Chi Chang-yan ab , Zhang Ji-xian a , Liu Zheng-jun a 
a Key Laboratory of Mapping from Space of State Bureau of Surveying and Mapping, Chinese Academy of Surveying 
and Mapping, 
Beijing, 100039,P.R.China - bingyu0122@163.com 
b College of surveying and Mapping, Shandong University of Science and Technology, Qingdao, 266510, P.R.China. 
KEY WORDS: Stripped Noise; Image Processing; Frequency Spectrum; MSE; PSNR. 
ABSTRACT: 
Noise is an important factor that influences image quality, which is mainly produced in the processes of image acquirement and 
transmission. Noise reduction is necessary for us to do image processing and image interpretation so as to acquire useful information 
that we want. Because the working status of image transmitter is influenced by varied of factors, such as the environment of image 
acquired, different noises can be dealt with in different ways. Through the analysis of image spectrum, its difference can help us to 
choose different methods to do noise reduction while the information of the image is reduced to be the most. This paper illustrates 
some methods of noise reduction and takes one test image as an example. Since this image was affected by notable striping, the noise 
reduction methods of this stripped image are mainly studied. Grounded on the previous studies by some experts, we have applied 
some algorithms to this image. The Gray Value Substitution and Wavelet Transformation are satisfactory in stripped noise reduction. 
Then, MSR and PSNR are calculated to evaluate the processed image. Results suggest that the methods used in this paper are 
suitable in processing this noise. 
1. INTRODUCTION 
Stripped noise is a usual phenomenon both in multi-sensor and 
sole sensor on the satellite and airplane platform. Because the 
working status of image transmitter is influenced by varied of 
factors, such as the environment of image acquired, different 
noise can happen in different ways. Striping is an important 
factor that influences image quality acquired by linear array 
CCD blocks. This may be more crucial for spectrometers 
because of the imperfect calibration of the detector 
characteristics and the necessity of higher CCD quality, which 
results in the most common striping [1] . Noise reduction is an 
important and basic part in remote sense image processing. Not 
only spatial noise but also spectral noise may exist in the image 
because of the influnece of natural light, surface topography, 
mixed pixel, etc. there are mainly two kinds of noise. One is 
plus noise and the other is multiply noise. Plus noise is usually 
caused by the noisy source and overlaid on the image to be 
shown: 
y(u, v) = x(u, v) + n(u, v) 
Where x(u,v) = prior image 
y(u.v)= the noisy image 
n(u,v)= the noise 
Multiply noise has the ablity of modulability to the image, when 
the gray value changes little and noise is small, multiply noise 
can be taken as plus noise approximatly. 
y(u,v) = x(u,v) + x(u,v)n(u,v) ^ 
two ways to reduce the noises. One kind is space enhancement 
and the other is frequency enhancement. 
Image frequency spectrum can express the characteristics of 
image noise. Through the analysis of image spectrum, its 
difference can help us to choose different methods to do noise 
reduction while the information of the image is reduced to be 
the most. The spectrum has been done by Fourier 
Transformation to give a definite visual show. 
This paper discusses the methods of noise reduction of this 
stripped image. Grounded on the previous studies by some 
experts, a test image is given and these methods are used to do 
stripped noise reduction. Then, we calculate the MSR and 
PSNR to evaluate the processes image. At the end of the paper, 
the comparisons of different noise reduction methods. 
2. METHODS OF NOISE REDUCTION 
Now, there are many methods to reduce noise. Trational median 
filter and mean filter are used to reduce salt-pepper noise and 
Gauss noise respectively. But when these two noises exist in the 
image at the same time, using only one filter method can not 
achive the wanted result. Suppose every sensor has the same 
balanced radiation distribution, the subimage histgram of every 
sensor is adjusted by histgram adustment to one referrence 
adjustment to realize noise redution. The precondition of this 
method has much limit and not use when the involuted surface 
contains different objects. Besides, it is suitble in the image 
after not before the geometrical rectification. Principal 
component analysis (PCA) changes the noise PC image value to 
constant then the result image is got by inverse transformation. 
However, striped noise is difficult to reduce because the noise 
usually exists in the PC image, and the computation needs much 
time. Low pass filter in frequency domain is suiltable to remove 
high frequency noise. 
Noise is plus noise in most cases. Moreover, there are uaually