采用3D Gaussian Facet模型的亚体素表面检测

Surface Detection with Subvoxel Accuracy Using 3D Gaussian Facet Model

提出一种基于Gaussian facet模型的3D边缘检测算法.首先利用Gaussian加权最小二乘拟合,引入空间权因子表达图像采样点对模型参数估计的相对重要度,扩展了经典Haralick facet模型,建立了3D Gaussian facet模型及其计算公式;然后采用抗噪性好的3DIDDG算子估计梯度方向,并在该梯度方向上计算二阶方向导数过零点,以获得表面点亚体素位置.实验结果表明,该算法能有效地降低邻近边缘干涉对检测结果的影响,可更好地提取尺寸较小的结构边缘.

This paper presents a 3D edge detection algorithm based on Gaussian facet model. First we generalize the classic Haralick facet model and introduce the 3D Gaussian facet model using the Gaussian weighted least squares fitting, which uses spatial weights to express the relative importance of image samples in estimating model parameters. Then we employ the 3D integrated directional derivative gradient （IDDG） operator to robustly estimate the gradient direction, and along this direction the zeros of the second directional derivatives are computed to locate the subvoxel positions of the surface points. Experimental results show our method can effectively reduce the interference of adjacent edge and can achieve good performance in extracting edge points of small structures.