Xu Haohao, Zhang Qi, Dong Huaipeng, Jiang Xiyuan, Shi Jun
Annu Int Conf IEEE Eng Med Biol Soc. 2018 Jul;2018:874-877. doi: 10.1109/EMBC.2018.8512314.
Speckle noise corrupts medical ultrasound images and suppression of speckle noise is valuable for image interpretation. This paper presents a new method for speckle suppression named the maximum likelihood based weighted nuclear norm minimization (MLWNNM) filtering by integrating the maximum likelihood estimation (MLE) with the weighted nuclear norm minimization (WNNM). The MLE is first used to get an initially filtered image with reduced Rayleigh distributed noise, and then the WNNM is applied to further improve the denoising effect by preserving and enhancing tissue details. Simulation work shows that when the noise variance is as high as 0.14, the MLWNNM improves the Pratt's figure of merit, peak signal to noise ratio, and mean structural similarity by 123.51%, 0.84%, and 6.13%, respectively, in contrast to the best values of other six methods. Experimental results on clinical ultrasound images suggest that the MLWNNM outperforms other six methods in noise reduction and detail preservation.
散斑噪声会破坏医学超声图像，抑制散斑噪声对于图像解读很有价值。本文提出了一种新的散斑抑制方法，即基于最大似然估计的加权核范数最小化（MLWNNM）滤波，该方法通过将最大似然估计（MLE）与加权核范数最小化（WNNM）相结合。首先使用最大似然估计来获得一个初始滤波图像，该图像的瑞利分布噪声有所减少，然后应用加权核范数最小化来通过保留和增强组织细节进一步提高去噪效果。仿真工作表明，当噪声方差高达0.14时，与其他六种方法的最佳值相比，MLWNNM分别将普拉特优值、峰值信噪比和平均结构相似性提高了123.51%、0.84%和6.13%。临床超声图像的实验结果表明，MLWNNM在降噪和细节保留方面优于其他六种方法。
