军事知识服务平台
An image denoising method for the Chinese first polar microsatellite(Ice Pathfinder)imagery
摘要 Strip noise is normally contained in optical remote sensing images from satellites,which impacts on their scientific applications.The"Ice Pathfinder"(code:BNU-1)is China's first small satellite designed for monitoring polar climate and environmental changes.Although radiometric calibration and correction have been carried out for all BNU-1 multis-pectral remote sensing images,strip noises are still inevitable in the images.The BNU-1 multispectral remote sensing images contain two types of strip noises.One is narrow-bandwidth strip noises with high-frequency appearances.Another type is wide-bandwidth strip noises with low-frequency appearances.This study develops an image denoising method to remove strip noises in BNU-1 images.The proposed method uses wavelet transform to decompose the BNU-1 multispectral remote sensing images into the low-frequency and high-frequency components.For the low-frequency components,a linear relationship between digital number(DN)values of the non-noise pixels in the BNU-1 multis-pectral images and the DN values of the corresponding pixels in the BNU-1 panchromatic band images is established to correct the DN values of the noise pixels in the multispectral images.For the high-frequency components,the method applies a threshold method to remove the strip noises.Finally,the inverse wavelet transform is used to restore the denoised BNU-1 multispectral remote sensing images.The proposed method is applied to 13 BNU-1 multispectral remote sensing images obtained from May to August in 2020 and 2021.The wavelet transform,the Savitzky-Golay(SG)filtering,and the Gaussian filtering are also utilized to remove the strip noises in the BNU-1 images for comparing with the proposed method.The results indicate that the Inverse Coefficients of Variation(ICV)and the Improvement Factors(IF)of the denoised image obtained by the proposed method are 47.21 and 23.90,which are significantly higher than the wavelet transform and SG filtering,and close to the Gaussian filtering.However,the visual performance of the BNU-1 multispectral images after denoising by the proposed method is better than that of Gaussian filtering.
刊 名 海洋学报(英文版) ISTICSCI
年,卷(期): 2026, 45(7)
机标分类号: X913.4 TU247 H315.9
在线出版日期: 2026年10月20日