基于双树复小波变换与稀疏表示的牙隐裂OCT三维图像融合  

OCT three dimensional fusion of tooth cracks based on DTCWT and sparse representation

作　　者：石博雅[1] 董潇阳 SHI Boya;DONG Xiaoyang(School of Electronics and Information Engineering,Tiangong University,Tianjin 300387,China)

机构地区：[1]天津工业大学电子与信息工程学院,天津300387

出　　处：《天津工业大学学报》2025年第1期62-68,共7页Journal of Tiangong University

基　　金：天津市自然科学基金项目(19JCYBJC16200);国家自然科学基金项目(61903273);天津市科技计划项目(20YDTPJC01090)。

摘　　要：针对采用光学相干层析(OCT)技术进行体积较大的前磨牙和磨牙的隐裂检测时,仅从单一扫描视角采集可能存在误检或漏检的问题,提出一种双树复小波变换(DTCWT)与稀疏表示(SR)相结合的牙隐裂三维图像融合方法。利用扫频OCT对人工牙隐裂模型从2个扫描视角进行成像,经过三维图像配准后,利用双树复小波变换对图像进行分解。对于低频子带进行稀疏表示,采用“最大L1范数”规则进行融合,高频子带采用“绝对最大”规则融合,最后通过DTCWT重构得到融合后的图像。实验结果表明:采用本文方法融合后的牙隐裂图像可以得到裂纹的完整信息,获得准确的定位和分级,各方面性能均优于单独采用各多尺度分解方法和稀疏表示方法,标准差(SD)、平均梯度(AG)、空间频率(SF)和边缘信息评价因子(Q)的值分别平均提高到36.7、6.0、27.9和0.74,有效提高了OCT牙隐裂检测的准确性。When optical coherence tomography(OCT)is used to detect tooth creaks in large anterior molars and molars,the problem of false detection or missed detection may exist only from a single scanning viewpoint acquisition.A dual-tree complex wavelet transform(DTCWT)combined with sparse representation for OCT 3D image fusion of tooth creaks was proposed to solve the problems.The artificial tooth-cracked models were imaged from two scanning views using Swept Source OCT,and after 3D image alignment,the images were decomposed using DTCWT.The low-pass bands were fused by sparse representation using the“max-L1”rule,while the high-pass bands were fused using the“max-absolute”rule,and the fused images were finally reconstructed by DTCWT.The experimental results show that the fused tooth creaks images can obtain the complete information of cracks,so that the cracks can be accurately localized and graded.This algorithm outperforms each multi-scale decomposition method or the sparse representation method alone in all aspects,the values of SD,AG,SF and Q increased to 36.7,6.0,27.9 and 0.74,respectively,effectively improving the accuracy of OCT tooth creaks detection.

关 键 词：牙隐裂 光学相干层析 稀疏表示 双树复小波变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]