基于预滤波和子块排序的稀疏域SAR图像去噪

Prefiltering and Patch Ordering-based SAR Image Despeckling via Sparse Domain Filtering

为了抑制合成孔径雷达(SAR)图像中存在的相干斑噪声,从而得到高质量的SAR图像,文中通过在子块排序之前引入预滤波处理提出了一种新的SAR图像噪声抑制方法。首先将SAR图像作对数变换;然后在子块排序前进行预滤波处理去除不相似的子块,从而可以避免不必要的计算,最重要的是能够抑制人为引入的纹理;最后,通过稀疏表示对排序后的子块进行滤波,对滤波后的子块进行逆排序、子图像平均和指数变换最终得到去噪后的SAR图像。仿真图像和真实SAR图像的实验结果表明文中提出的去噪方法的去噪效果评价指标峰值信噪比(PSNR)、结构相似度(SSIM)和等效视数(ENL)都达到了现有优良去噪算法的水平,去噪后的图像看不到明显的斑点噪声,图像清晰且纹理保存较好,均匀区域平滑。该方法具有良好的相干斑抑制能力和细节保留能力。

In order to suppress speckle noise in synthetic aperture radar(SAR) image and obtain high quality SAR image, this paper proposes a novel method of SAR image denoising by prefiltering the patches before patch ordering. This method firstly performs logarithmic transformation on the SAR image. Then, some dissimilar patches are removed by prefilter before patch ordering. It can avoid unnecessary calculations, suppressing the introduced artificial texture. Finally, ordered patches are filtered by sparse representation. The final denoised image can be reconstructed from the filtered patches via inverse permutation, subimage averaging and exponential transformation. The experimental results of simulated images and real SAR images show that the denoising method proposed in this paper reaches the level of the existing good denoising algorithms in terms of peak signal to noise ratio(PSNR)、structural similarity index(SSIM) and equivalent numbers of looks(ENL). The denoised image has reached the level of the existing good algorithms in visual effect. This method can well maintain the image details and at the same time suppress speckle noise in SAR image.