改进快速自适应双边滤波的SAR图像去噪

Improved Fast Adaptive Bilateral Filtering for SAR Image Denoising

针对已有算法在去除散斑噪声时存在边缘信息丢失的现象,提出一种基于平稳小波变换,结合快速自适应双边滤波、小波阈值去噪和优化贝叶斯非局部均值的算法(SFOBNLM算法),在保持边缘信息的同时有效地去除SAR图像中的散斑噪声.先用平稳小波变换将图像分解为近似子带和细节子带,对这两个子带分别进行快速自适应双边滤波和小波阈值处理;然后用小波逆变换对去噪后的图像进行重构;最后用OBNLM算法对图像进一步去噪.实验结果表明,SFOBNLM算法在峰值信噪比(PSNR)、结构相似度(SSIM)、特征相似度(FSIM)、边缘保持因子(EPF)、等效视数(ENL)和视觉质量方面均优于传统的滤波方法.

Aiming at the phenomenon of loss of edge information when removing speckle noise in existing algorithms,an algorithm based on stationary wavelet transform,combining fast adaptive bilateral filtering,wavelet threshold denoising and optimized Bayesian non-local mean(SFOBNLM algorithm)was proposed,which effectively removes speckle noise in SAR images while preserving edge information.First,the image was decomposed into approximate sub-bands and detail sub-bands by stationary wavelet transform,and fast adaptive bilateral filtering and wavelet threshold processing were performed on these two sub-bands respectively;then the denoised image was reconstructed by inverse wavelet transform;finally,the image was further denoised with the OBNLM algorithm.The experimental results show that the SFOBNLM algorithm is superior to traditional filtering methods in terms of peak signal-to-noise ratio(PSNR),structural similarity(SSIM),feature similarity(FSIM),edge preservation factor(EPF),equivalent view number(ENL)and visual quality.