Based on the structural tensor of projection, this study aims to address and test a new improved algorithm applying to the distort projection data to generate a high qualified image by reducing the artifacts and noise from scattering in the cone-beam computed tomography (CBCT). Since the scattering information has a large relationship with the structure of the object, which is reflected by the projection, regional model knowledge for scattering is accomplished by finding the relationship between projection and scattering. As the tensor, the gradient of projection is first calculated in the process for estimating the direction and structural edge of the object. Then, the Determinant and Traces of the tensor map with different characteristics are computed to determine the different regions. By modeling and fitting the regions of scattering distribution, the knowledge of scattering parameters corresponding to a different region is obtained. Based on the similarity of scattering distribution in adjacent angles, the scatterings with angle sequence are completed by interpolating the prior knowledge obtained through the sparse sampling. By performing the studies on polychromatic X-ray to test the performance of the scattering estimation algorithm, the results show a significant improvement in the images that are reconstructed from the corrected projection. The root mean square error (RMSE) of the proposed method is reduced by 21.8% and 39.8%, respectively. Peak signal to noise ratio (PSNR), and universal quality index (UQI) also indicate better uniformity, where the PSNR is increased by 7.4% and 56.7%, UQI is increased by 70.8% and 262.3% for experimental #Wheel and #Cylinder, respectively.

中文翻译：

---

基于投影结构张量和建模的锥束CT角序列散射测量和估计。

基于投影的结构张量，本研究旨在解决并测试一种新的改进算法，该算法可通过减少锥束计算机断层扫描（CBCT）中的散射造成的伪影和噪声，将其应用于失真投影数据以生成高质量图像。由于散射信息与物体的结构有很大的关系，这由投影反映出来，因此散射的区域模型知识是通过找到投影与散射之间的关系来实现的。作为张量，首先在估计对象的方向和结构边缘的过程中计算投影的梯度。然后，计算具有不同特征的张量图的行列式和迹线，以确定不同的区域。通过对散射分布的区域进行建模和拟合，可以获得与不同区域相对应的散射参数的知识。基于相邻角度散射分布的相似性，通过对通过稀疏采样获得的先验知识进行插值，可以完成具有角度序列的散射。通过对多色X射线进行研究以测试散射估计算法的性能，结果显示了从校正投影重建的图像的显着改善。该方法的均方根误差（RMSE）分别降低了21.8％和39.8％。峰值信噪比（PSNR）和通用质量指数（UQI）也表明更好的均匀性，其中PSNR分别增加7.4％和56.7％，UQI分别增加70.8％和262。