基于视觉模型的SAR图像压缩

HVS-based SAR Image Compression

针对SAR图像压缩方法通常反利用小波变换去除图像数据间的统计冗余,而忽略图像的视觉冗余的问题,在讨论人类视觉系统(HVS)特性基础上,结合对比敏感度函数(CSF),对采用提升方案进行小波分解后的不同频带小波系数加权,然后采用多级树集合划分(SPIHT)算法形成嵌入式小波编码方法.实验结果表明,与标准SPIHT压缩方法相比,相同压缩比下,本算法重建图像客观评价标准几乎保持不变,且很好地保留了图像边缘和纹理信息,主观视觉效果有明显改善.

To remove the neglect of visual redundancy due to the SAR image compression methods based on wavelet transform, a new method based on human visual system （HVS） is put foward in this paper. First, SAS image is decomposed by lifting scheme, then wavelet coefficients in different subbans are weighted by the peak of the contrase sensitivity function （CSF） curve in the corresponding frequency band. At last, set partitioning in hierarchical trees （SPIHI） algorithm is used to code the wavelet coefficients to form the embedded bit stream. The results show that this method gets better subject visual quality under the same compression ratio with the equivalent objective evaluation results compared with SPIHT.