基于多小波的SAR图像去噪与压缩

SAR Image Compression Based on Multiwavelet Combining with Speckle Noise Reduction

SAR图像固有的乘性相干斑噪声降低了图像的相关性,增加了信息熵,影响了图像压缩的性能。多小波能够同时拥有正交性、紧支性和对称性,比单小波具有更多的自由度。因此提出了在多小波域进行去噪和压缩相结合的SAR图像编码算法。首先对图像进行多小波变换,采用改进的软阈值法抑制相干斑噪声同时对图像边缘进行保护,再对多小波系数重排建立空间方向树,然后采用多级树集合划分(SPIHT)算法进行编码。实验结果表明,该算法改进了重建SAR图像的PSNR,同时对相干斑噪声进行了有效的抑制。

Synthetic aperture radar （SAR） images are corrupted by multiplicative speckle noise which decreases correlation between adjacent pixels, increases the information entropy and limits the performance of the classical coder/decoder algorithms in spatial domain. The relatively new transform of multiwavelet can possess desirable features simultaneously, such as short support, orthogonality and symmetry, while scalar wavelets cannot. Thus a compression scheme combining with speckle noise reduction within the multiwavelet framework was proposed. After multiwavelet transform, modified soft-thresholding denoising method was applied to reduce speckle noise which protected more edge information at the meantime. Then multiwavelet coefficients were rearranged to reconstruct spatial orientation tree and coded by set partitioning in hierarchical trees （SPIHT） algorithm. Experimental results show this coding method achieves favorable peak signal to noise ratio （PSNR） and superior speckle noise reduction performances.