摘要
条带噪声的存在不但妨碍高光谱图像的目视判读,而且制约高光谱遥感的定量应用。针对小波变换法条带噪声去除过程中遇到的条带噪声和图像有用信息难以有效分离的问题,根据小波变换的方向性和数学显微镜特性,提出了一种新的基于小波变换的条带噪声去除方法。这种方法首先对含有条带噪声的图像进行一定层数的小波分解;然后对每一层分解得到的与条带噪声分布方向相同的子图像再进行一定层数的小波分解,从而实现条带噪声和图像有用信息的有效分离,将含有条带噪声的子图像置零;最后利用小波反变换得到去除条带噪声的图像。以欧洲空间局PROBA卫星上搭载的CHRIS高光谱数据为例,采用相关系数(R)、结构相似度(SSIM)和峰值信噪比(PSNR)3个定量指标,对比分析了新方法与矩匹配法、傅立叶滤波法和小波阈值法的条带噪声去除效果。结果表明新方法去噪后的图像具有最高的R、SSIM和PSNR,新方法能够有效地去除高光谱图像中的条带噪声,同时较好地保留了原始图像的有用信息。
Stripe noise,which not only hinders the visual interpretation of hyperspectral images,and restricts quantitative application of hyperspectral remote sensing.The commonly used wavelet transform filtering methods have difficulty in separating stripe noise and useful information of the original images.For this reason,with reference to the characteristic of directivity and "mathematic microscope"of wavelet transform,a new destriping method based on wavelet transform has been proposed In this study.Firstly,this method decomposes the original images containing stripe by a certain scale of wavelet transform,then the high frequency sub-images of every scale which have the same direction with stripe noise and the low frequency sub-images are decomposed separately by wavelet transform to a degree where stripe noise and useful information can be effectively identified.Consequently,the values of sub-images serious corrupted by stripe noise are set to zero.Finally,the destriped images are constructed by inverse wavelet transform.We take CHRIS hyperspectral data from European Space Agency satellite PROBA as an example.The proposed method is compared with moment matching,Fourier filtering method and wavelet threshold method by quantitatively evaluating those four destriping methods using three indices:correlation coefficient(R),Structure Similarity(SSIM)and Peak Signal to Noise Ratio(PSNR).The results show that the image processed by the new method has the highest R,SSIM and PSNR,Which indicate the method proposed in this study effectively eliminates stripe noise in hyperspectral image while preserving the useful information of the original image.
出处
《遥感技术与应用》
CSCD
北大核心
2015年第6期1168-1175,共8页
Remote Sensing Technology and Application
基金
国土资源部公益性行业科研专项(201011019-07)
关键词
高光谱图像
条带噪声
小波变换
条带分离
条带去除
hyperspectral image
Stripe noise
Wavelet transform
Stripe separation
Destriping