结合分数阶微分与结构张量的医学图像细微结构增强

Enhancement of Medical Image's Detail Structure Based on the Combination of Fractional Differential with Structure Tensor

从分数阶微分和结构张量的特点出发,考虑到分数阶微分算子具有较好的纹理细节表达能力,结构张量具有保持图像几何形状的特性,本文通过构造分数阶结构张量,提出了结合分数阶微分算子与结构张量的医学图像细微结构增强方法.分别针对计算机断层扫描(CT)图像和磁共振(MR)图像做了增强实验,结合小波分解,对低频信号作Tiansi微分算子锐化以加强轮廓提取,对高频信号构建基于分数阶结构张量的特征值相干性度量,并根据此相干性度量加权高频信号以得到更清晰连续的纹理细节.实验结果表明,结合分数阶微分算子与结构张量的增强方法对图像的纹理细节表征能力更强,增强后的图像信息更为丰富,细节更为清晰,从而具有较好的临床应用潜力.

Starting from the characteristics of fractional differential and structure tensor,considering that fractional differential operator has good texture detail expression and structure tensor has the property of preserving the image geometry,we proposes a medical image＇s detail structure enhancement method based on the combination of fractional differential operator with structure tensor by constructing fractional differential-based structure tensor. Image enhancement experiments are carried out with computed tomography images and magnetic resonance images respectively. Moreover,combining with the wavelet decomposition,for the low-frequency signal, we use Tiansi differential operators for sharpening to enhance the contour extraction, for the high-frequency signals, we construct the eigenvalue coherence based on fractional differential-based structure tensor, then according to this coherence, weighing the high-frequency signals to obtain clearer and continuous texture details. To conclude,experimental comparison results show that the combination of fractional differential operator and structural tensor enhancement method can enrich the texture details of the image, which can provide more information and clearer details and has better clinical application potential.