1271 
THE INFLUENCE ON IMAGE QUALITY UNDER RELATIVE CALIBRATION 
ACCURACY 
Hu Yongfu*, Zhang Yufeng, Zong Xiaoying 
Beijing Institute of Space Mechanics & Electricity, 9201mailbox, Beijing, 100076, China -huyongfu@gmail.com 
Commission I, SS-4 
KEY WORDS: accuracy, analysis, calibration, imagery, method, pushbroom, remote sensing, 
ABSTRACT: 
Strips can be seen in raw remote sensing images of CCD camera. The object of relative radiometric calibration is to eliminate them. 
This paper discusses the influence on image quality under different calibration accuracy. The laboratory radiometric calibration data 
of the China Brazil Earth Resources Satellite (CBERS) 02B CCD camera were used to analyse the data distribution which has been 
processed through the relative radiometric correction. A simulation method based on the statistic and analysis of the corrected data 
was found and used to simulate the corrected result after validating. The influence on image quality under different relative 
calibration accuracy was simulated with the method and Peak Signal to Noise Ratio (PSNR) was used to evaluate the results. 
1 INTRODUCTION 
Relative radiometric correction is the first step of the remote 
sensing image processing. The theory of normalized coefficients 
algorithm, two-point algorithm and least-squares algorithm are 
usually used in relative radiometric calibration. According to 
the relative linearity curve (the linearity between the output DN 
value and the expected DN value of a pixel), the appropriate 
algorithm is selected to calculate the calibration coefficients and 
correct the raw remote sensing image. By this way, the strips in 
the images are reduced. But they can’t be eliminated absolutely 
because of the relative calibration error. The influence on image 
quality under different calibration accuracy was researched in 
this paper. 
The process can be divided into three steps. Firstly, laboratory 
radiometric calibration data of CBERS 02B CCD camera were 
processed and the relative calibration coefficients were obtained. 
Secondly, the calibration coefficients were used to correct the 
calibration image and the distribution of the corrected data was 
analyzed. Finally, the influence on image quality under different 
relative calibration accuracy was simulated with Matlab. 2 
2 ANALYSIS OF CBERS 02B CCD DATA 
1 
g 
Fig.l. Corrected result of spectrum B2 data 
After correction, the DN values varied in 2~3 and the standard 
deviation is less than 0.5 in a row. The result is good enough. 
The statistic histograms of the corrected data in all radiance 
levels are shown in Fig. 2. 
Original radiometric calibration data of CBERS 02B CCD 
camera were acquired by viewing an integrating semi-sphere in 
laboratory before the satellite was launched. These data are 
processed and analyzed in this section. 
Taking the data process of Band B2 as an example. The relative 
linearity curve is linear by analyzing the data. So the linear 
least-squares algorithm is selected to calculate the relative 
calibration coefficients and correct the raw data. The corrected 
result of B2 is shown in Fig. 1.