X射线数字纹理图像的特征提取

以减少图像背景和结构纹理对特征提取的干扰为目的，提出了先去除背景和纹理，再进行特征分析的算法.该算法通过最小二乘法则拟合了类抛物线曲面函数提取数字射线图像的背景，减少了背景对图像特征的模糊，在此基础上，针对图像的结构纹理特点确定几何分布参量，定义减法运算公式消隐图像纹理，减少了纹理造成的图像灰度起伏，图像灰度级分布均匀.在平坦的图像背景中，根据数字射线图像信号点服从正态分布规律特点，设定阈值进行特征分割.

The Characteristic Separation of the Textural X-ray Image

Subtracting the background of the image by the polynomial function based on the Least Square-Fit was put forward. The influence of the background being on the characteristic was reduced by the subtracting. Then, according to the structure of the workpiece, the texture was first removed in order to make the gray level even. At last, following normal distribution of the signal of the digital image, a threshold was set to make the characteristic apart.