【作者】 赵英亮；

【导师】 韩焱；

【作者基本信息】 中北大学 ， 信息处理与重建， 2010， 博士

【摘要】 论文根据三维CT技术的发展现状和国内外对三维CT技术提出的需求,依赖被检测对象内部结构特征的先验知识和模型,在对射线图像增强及散射恢复等预处理技术研究的基础上,结合FDK重建算法,研究直接重建特征的三维高分辨率CT重建技术。特征CT重建的质量取决于投影图像特征的提取,然而从锥束CT系统的成像原理可知,许多因素会影响CT投影图像的性能。在对锥束数据获取和扫描方式研究的基础上,提出了一种基于区域连通修正H参数的分形图像增强算法。该算法首先建立投影信号的分形模型,计算H参数及分形维；之后,针对分形算法对噪声敏感性问题,利用滑动窗口内各点与中心点连通关系,对其H参数进行修正,对图像进行特征增强。实验结果表明,本方法既有效地抵抗了噪声的影响,同时又凸显了射线图像中灰度差别较小的目标边缘。利用偏振分析,研究了射线散射的影响,提出一种基于ICA的盲估计分离散射算法。该方法通过建立散射的ICA模型,对小波分解后高频域的子带图像,利用最小互信息熵做为盲估计的准则,推导出其优化的估计量。实验结果表明,该算法对投影图像因散射引起的模糊具有较好的恢复效果。在分析研究传统三维CT重建算法的基础上,以圆轨迹扫描方式为背景,结合FDK算法,提出一种三维特征重建算法,并从理论的角度进行了推导。基于特征重建的CT算法基本理论,结合小波变换在RADON变换前后的特性,对小波应用于三维CT重建算法进行了理论与实现的研究工作,利用二维小波的边缘检测方法实现结构特征提取,并将其所提取的特征应用于三维重建中,得到所感兴趣的重建结果。从通用性角度与多特征重建出发,将EMD引入三维特征CT重建算法并进行了理论推导,提出了基于信息熵结合PSNR的EMD最优分解控制方法,进一步通过实验验证了方法的有效性。实验结果分析表明,基于EMD的CT重建方法能够有效的重建出原始图像的多种特征信息。最后将三维特征CT算法与相关算法进行参数的比较,证明其具有高频保留好,抗噪能力强,滤波器易实现及运行速度快的优点。更多还原

【Abstract】 According to the development of three-dimensional CT Technology and the domestic and abroad demands, relying on the prior knowledge and models of the internal structure of the object, on the study of x-ray image enhancement and scattering restoration, combined with FDK reconstruction algorithm, the three-dimensional high resolution CT reconstruction method to directly reconstruct the characteristics was studied.In the CT reconstruction, the quality depends on the feature extraction of projection images, but many factors will affect their performance from the imaging principle of cone beam CT system.On the introduction of the cone beam data acquisition and scanning, a fractal image enhancement method based on the regional connectivity amendment parameter was raised. The fractal model was firstly constructed and H parameters and fractal dimensions were computed. Considering its sensitivity to noise, the connectivity relationship between the center and ambient gray scale inner the slider windows was used to amend H parameters respectively, and further enhance the image for better results. The simulation results showed that the method can not only stress the edge of smaller gray-ray difference but also be of better noise immunity.The polarization analyzation was used to study the scattering and a new restoration method based on the blind separation was raised. An ICA (Independent Component Analysis) model was firstly constituted based on the radiograph mechanism and the sub-band images in high frequency domain by wavelet transform were utilized as estimating source. We utilized the least mutual information entropy as the criteria for blind separation and deduced to the optimization estimation. The experimental results showed that the method can effectively depress the scattering and improve the contrast. On the analyzing of the traditional three-dimensional CT reconstruction, based on the circular scanning, combined with FDK algorithm, a new 3D industrial CT reconstruction algorithm to directly reconstruct the characteristics was brought forward, and the feasibility of the new algorithm was theoretically deduced. Using the fundament of the new algorithm, contacted with the traint of the wavelet before and after RADON, it was deduced to use 2-D wavelet transform to extract the characteristics, and the characteristics were directly reconstruct in the 3D CT to obtain the interested parts. From the generability and multi-characteristics reconstruction EMD was applied in the theory and implementation of three-dimensional CT reconstruction. And an optimal EMD decomposing control method was proposed based on the combination of information entropy and PSNR, and further experimentally verified. The experiment results showed that the CT reconstruction method based on EMD can effectively reconstruct a variety of characteristic information of the original image. By the comparison with the related algorithms, the new method is proved of better preservation to the high frequency information, stronger resistance to noise, easier filter design and shorter time-consuming.更多还原