用于3D生物医学图象边缘检出的一种新零交叉算法

在三维生物医学图象分割中，Marr算法的缺点是当模板较大时运算速度很慢。Haralick算法的缺点是曲面拟合系数和方向导数的方向较难确定。对此，作者提出了一种新的零交叉算法。它是先对图象进行对称二阶微分运算，然后由平面拟合检出过零点。其优点是模板较大时运算速度比Marr算法快，且采用平面拟合具有较好的抗干扰性。

A Novel Method of Zero Cross for 3D Biomedical Image Edge Detection

In image processing and recognition, zero crossing is a good method for image edge detection. The common methods of zero crossing are Marr and Haralick algorithms. In 3D biomedical image segmentation,the disadvantage of Marr is low speed when a large kernel is used. The disadvantage of Haralick is the difficulty for determining the coefficients of the surface fitting and the orientation of the directional second order derivation. For this reason, the authors propose a new method of zero crossing. In this method, a symmetric second order derivation is first performed. Then a plane fitting is performed to detect zero crossing pixels. The advantage of our method is that the speed is faster than Marr when the kernel is large and the plane fitting can reduce noise.