TDI-CCD遥感图像条带噪声的消除

针对TDI-CCD遥感图像固有条带噪声的灰度值在原始信息中变化比较缓慢的特 点及小波变换对奇异点检测的优越性, 提出了一种快速傅里叶变换和小波变换 相结合的方法来确定条带噪声频率点所在的位置, 采用改进的陷波滤波器对条 带噪声的频率成分进行消除. 另外, 提出一种基于提升小波变换的改进阈值 函数法, 用以消除图像的条带噪声. 实验结果表明, 在消除条带噪声方面, 改 进的陷波法优于改进的阈值函数法, 而改进的阈值函数法优于硬阈值函数和软 阈值函数法. 利用改进的陷波法消除条带噪声后, 图像的质量得到提高, 其PSNR达到46.4181dB, NMSE减小到0.00007, 相对基于提升小波变换的 几种阈值函数法, PSNR提高了3～4dB, 而NMSE则减小了 0.00007～0.00011. 本文提出的方法能较好地消除遥感图像的条带噪声, 同时能够较好地保持原图像的特征, 具有可行性. 

Destriping for TDI-CCD Remote Sensing Images

Based on the characteristic of striping noise in TDI-CCD images,a new destriping noise technique which combines FFT with wavelet transform to search the frequency point of the striping noise is presented.The improved notch filter is selected to eliminate the striping noise. In addition,an improved threshold function based on lifting wavelet transform is used to eliminate the stripes.Experiment shows that the method of improved notch filter is superior to the improved threshold function while the improved threshold function is superior to the hard and the soft threshold function.After destriping using the improved notch filter,the quality of image is enhanced.The PSNR of the destriping image processed by improved notch filter is increased to 46.4181 dB and the NMSE is decreased to 0.00007.Compared to the methods of threshold function based on lifting wavelet transform,the PSNR of destriped image is improved by 3~4 dB and the NMSE is reduced by 0.000 07~0.00011.The method of improved notch filter which combined FFT with wavelet transform and the methods of threshold function based on lifting wavelet transform can eliminate the striping noise while preserve the characteristic of the original image,so the several methods are of feasibility.