基于NSCT的超声图像自适应阈值去噪方法

Denoising Method of the Ultrasonic Image Adaptive Threshold Based on Nonsubsampled Contourlet Transform

根据超声图像乘性斑点噪声模型以及非下采样Contourlet子带系数的统计特性,提出了一种改进的基于非下采样Contourlet变换(NSCT)的斑点噪声抑制算法。上述算法通过把对数变换后的超声图像的NSCT子带系数模型化为广义高斯分布(GGD),给出一个改进的自适应Bayes Shrink阈值,用软阈值函数处理来实现超声图像的去噪处理。实验结果表明,相比于已有的基于小波和基于NSCT的Bayes Shrink阈值去噪法,在去噪性能和视觉效果上,上述算法都有明显的提高和改善,并且也有效地保持了原始图像的边缘和细节信息。

Combining the ultrasonic image multiplicative speckle noise model and the statistic characteristic of the non-down-sampling Contourlet sub-band coefficient,an improved speckle noise suppression algorithm based on non-down-sampling Countourlet Transfor-mation（ NSCT） is presented in this paper. The ultra-sonic image NSCT subband coefficient model after the logarithm transformation is transformed into the General Gaussian Distribution（ GGD）by this algorithm. An improved adaptive Bayes Shrink threshold value is given and the noise Suppression processing is achieved by a soft threshold function. The test result shows,compared with the existing noise suppression methods based on Wavelet Transform and Contourlet Transform,the algorithm presented in the paper has evident improvement in the performance evaluation of noise suppression and the subjective visual effect. Meanwhile,the marginal and detailed information of the original image are maintained effectively as well.