基于变分法的超声图像斑点噪声自适应滤波算法  

An Adaptive Benoising Algorithm for Ultrasound Image Based on Variations

作　　者：张琼[1,2] 沈民奋[2] 方若宇[2] 

机构地区：[1]中国科学技术大学电子科学技术系,安徽合肥230026 [2]广东省汕头大学工学院,广东汕头515063

出　　处：《北京生物医学工程》2011年第3期253-258,共6页Beijing Biomedical Engineering

基　　金：国家自然科学基金(N60571066);汕头大学青年科研基金(YR10005)资助

摘　　要：各向异性扩散模型在去除超声图像斑点噪声时不能有效保护图像细节,针对上述问题本文提出基于变分法的自适应最小能量去噪模型。首先直接将由微分方程表示的各向异性扩散模型转化为最小能量变分模型;然后引入欧拉弹性能量模型,在去除噪声的同时有效地保护和增强图像细节。同时为了解决数值求解过程中出现的迭代次数与迭代步长的矛盾,本文还提出迭代停止准则和自适应变步长去噪算法。仿真和真实超声图像的实验结果表明基于变分法的超声图像斑点噪声自适应滤波算法在去噪的同时能够很好地保护细节信息,而且能有效地减少迭代次数。In order to solve the problem that speckle reducing anisotropic diffusion model can not effectively preserve ultrasound image edges while filtering, an adaptive smallest energy denoising algorithm based on variations is proposed. First, a minimum-energy variation model is derived from anisotropic diffusion differential equation. And the Euler＇ s elastic energy model is introduced to enhance the ability of edgepreservation while filtering. Then in order to solve the conflict between iteration times and step sizes, an iteration-stopping criterion and an adaptive step-size iteration speckle removal scheme are proposed. To evaluate the proposed algorithm, both synthetically introduced speckle images and real medical ultrasound are used. The simulation results show that the proposed method not only preserves image details effectively while filtering, but also reduces the iteration times.

关 键 词：变分法 斑点噪声去除 边界保护和增强 欧拉弹性模型 自适应变步长迭代 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置] R318.04[自动化与计算机技术—控制科学与工程]