• 论文与研究报告 •

### 基于X射线木材断层扫描反投影算法中的最优滤波器选取

1. 山东建筑大学信息与电气工程学院 济南 250101
• 收稿日期:2018-04-03 修回日期:2018-07-23 出版日期:2018-11-25 发布日期:2018-12-04
• 基金资助:
泰山学者优势特色学科人才团队（2015162）；山东建筑大学博士基金"基于X射线的木结构建筑用材无损检测系统研究"（XNBS1622）。

### Selection of the Optimum Filter in the Back Projection Algorithm Based on X-Ray Wood Tomography

Luo Rui, Ge Zhedong, Chen Longxian, Liu Chuanze, Zhou Yucheng, Xu Wenqi

1. School of Information and Electrical Engineering, Shandong Jianzhu Univercity Jinan 250101
• Received:2018-04-03 Revised:2018-07-23 Online:2018-11-25 Published:2018-12-04

Abstract: [Objective] In order to optimize the best filter in wood tomography image reconstruction and provide scientific methods for the study of wood tomography, the reconstructed image effects of R-L, S-L, Cosine, Hamming and Hamming filters commonly used in X-ray computed tomography scanning reconstruction algorithm were compared in this study, and the reconstructed image quality of filter function were also analyzed.[Method] According to the CT imaging principle, the image filtering process was analyzed, and the working characteristics of different filters and the design steps of the filters were discussed. Firstly, the projection data is stored in a binary two-dimensional array. The projection value of each point is determined by the probe point number and the rotation angle. The filter length is determined according to the length of the projection data sequence, and the projection data at different angles are zero-padded before and after, so that it constitutes a series of new projection data that is three times the ones of the original length. Then, a one-dimensional fast Fourier transform is performed, and the frequency domain result is multiplied by the filter frequency domain discrete form, and the projection data is filtered line by line. Finally, the filtered result of the product is inversely transformed by Fourier to obtain the projection data in the time domain. The imaging result of different filters are compared, and different filters are evaluated and analyzed.[Result] After comparing with the back-projected reconstruction image without filtering, the boundary of the growth ring of the five filtered images is clearer, the artifacts are less, and the edges are smoother. The image artifacts are directly reconstructed by the back projection without filtering with an extremely poor reconstruction effect, with the structural characteristics of the growth wheel and crack are not clear. After reconstruction with R-L, S-L and Cosine filters, respectively, the edges of the tomographic image are smooth and clear, and the non-directional periphery is diffused. After the Hamming and Hamming filters are reconstructed, respectively, the image artifacts are slight, and there are obvious corrugated artifacts. From the difference of each filtered image and the detection model, the image quality is the highest after reconstruced by the R-L filter, and the error of the Hamming filter is the largest after reconstruction.[Conclusion] The reconstructed image obtained by R-L filter has the smallest error, the reconstructed image is the clearest, the spatial resolution is high, and the artifact can be effectively removed. The filtered wood tomographic image can accurately determine the shape, position and size of the defect.