基于Hessian自适应增强的口腔结构光图像处理

Image Processing of Intraoral Structured Light Based on Hessian Adaptive Enhancement

为提升重建的牙齿三维点云的精度,解决口水、牙釉质等环境因素导致的结构光口腔扫描仪采集的条纹图像质量较差的问题,首先引入一种自适应中值广义总变分融合滤波方法来抑制噪声,并且有效保留图像细节信息;在此基础上提出一种基于Hessian矩阵特征值的自适应线性结构增强算法用于增强二维结构光图像中的条纹信息,算法根据统计的局部特征,自适应调整增强参数。实验结果表明,与传统Hessian增强方法相比,自适应Hessian增强后的条纹图像的熵增强函数(EMEE)和信息熵更高,得到的三维点云有更完整的细节特征和更高的匹配精度。

In order to improve the accuracy of three-dimensional tooth point cloud reconstruction and solve the problem of poor quality of fringe images collected by structured light intraoral scanner caused by environmental factors such as saliva and tooth enamel,a fusion filter of adaptive median and total generalized variation is proposed to suppress noise while effectively preserving the image details.Based on the smoothed image,an adaptive linear structure enhancement algorithm based on the eigenvalues of Hessian matrix is proposed to enhance fringe information in the two-dimensional structured light image.Compared with the traditional Hessian enhancement method,the experimental results show that the enhancement parameter can be adjusted adaptively according to statistical local features,the EMEE and entropy of stripe image enhanced by adaptive Hessian enhancement method are higher,and the 3D point cloud has more complete detail features and higher matching accuracy.