眼科高频超声成像中斑点噪声抑制算法的研究

摘要

目的 斑点噪声是超声图像中存在的固有问题,而在眼科高频超声这种更为精细的超声检查中,有效地抑制斑点噪声能提高图像的质量,有助于临床医生对病情的判别.方法 提出了一种新的基于拉普拉斯(Laplacian)金字塔的多尺度斑点去噪方法.采用Laplacian金字塔,从斑点噪声中分离出临床图像特征,根据每层子带图像不同尺度及特点,从小尺度到大尺度,首先采用改进后的八方向各向异性斑点去噪(SRAD)去除图像斑点,然后增强图像的边缘、细节及对比度等方面.该方法与传统的SRAD滤波及相干增强滤波(CEDIF)进行对比,采用等效视数及算法的时间耗费对实验结果进行量化评估.结果 与传统SRAD滤波及CEDIF滤波方法相比,基于Laplacian金字塔的多尺度各向异性斑点去噪方法均高于前两种方法(1.172 3 vs 1.122 3、0.929 3及0.864 0 vs 1.396 0、1.468 3).结论 本研究提出的基于Laplacian金字塔的多尺度各向异性斑点去噪方法在更有效地去除图像斑点噪声的同时,能很好地保存图像边缘及图像细节等.%Objective Speckle is the inherent problem exists in B-Mode ultrasound image,and effective speckle noise reduction will improve the image quality and contribute greatly to clinical doctors to diagnose,especially in fine ophthalmic examination with high frequency ultrasound.Methods This paper proposed a new speckle reduction method based on the Laplacian pyramid and multiscale analysis.In the Laplacian pyramid,true clinical features were separated from noise,according to the different bandpass ultrasound image characteristics in each layer,and the advanced eight directions speckle reduction anisotropic diffusion (SRAD)was adapted to suppressed the speckle noise,and the identified noise and the extracted features were selectively emphasized by suitable edge,coherence and contrast enhancement filtering from fine to coarse scales.The performance of the proposed method was compared to the traditional SRAD method and coherence-enhancing diffusion method by measuring the equivalent number of looks (ENL) and Time cost.Results The ENL and Time cost value of the proposed method were higher compared to the SRAD and the cedif method,i.e 1.172 3 vs 1.122 3,0.929 3 and 0.864 0 vs 1.396 0,1.468 3.Conclusions In summary,the proposed method can more effectively remove the speckle noise while preserving the edge and details of the images.