【作者】 刘宾；

【导师】 王黎明；

【作者基本信息】 中北大学 ， 信号与信息处理， 2014， 博士

【摘要】 X射线数字成像在检测具有大厚度比物体时，传统固定管电压成像模式易出现过曝光和欠曝光共存现象，导致图像中数据信息严重缺失，无法实现对检测对象结构信息的完整再现。目前人们通常采用厚度、密度补偿、局部透照、大电荷容量、高动态范围的光电成像器件来解决此问题，尽管增加透射剂量、增大了成本，但还是难以获得图像的完整可视效果。对此本文从系统控制设计、数据采集、图像融合及显示方面研究高动态范围X射线成像检测技术，旨在基于低动态范围成像器件，在有效控制物体厚度细分程度和相邻厚度区域灰度间的线性比例假设的基础上，对不同厚度区段对应的有效数据进行提取和融合，最终获取对应整个工件厚度范围的高动态范围射线图像，从而达到扩展成像系统动态范围的目的。针对融合高动态范围图像的显示问题，研究高动态范围图像的灰度压缩技术，使其在保证结构信息增强的同时，压缩融合图像的动态范围，使其能够在低动态范围显示设备上完整有效显示。本文提出了线性递变高压射线源控制的设计思想，该高压源能够实现按设定kV T曲线进行分段线性递变，同时配合探测器曝光时间间隔的调整，可以快速自动获取不同管电压间隔的透照图像序列，为后续有效信息提取和融合提供原始数据。通过对射线检测过程中的射线与物体的作用规律和灰度变化规律的研究，确定了相邻管电压下厚度重复率、图像灰度间线性比例关系、管电压间隔等参数的关联特性；提出了基于比例增益直方图的分析计算方法，可以对相邻管电压间的比例增益系数、线性程度和重复率等参数进行快速计算。提出了相邻管电压间隔图像对应厚度重复率约束的思想，能够实现对原始透照数据的二次抽样过程进行智能控制，并有效提取不同厚度区域对应的有效信息。通过比例增益直方图可以对相邻管电压对应有效信息的比例因子进行计算，通过比例因子对较低电压对应有效信息进行加权处理，当空间对应所有像素点均处理完毕后可以获得最终融合高动态范围图像，此种融合方法能够有效控制一定厚度变化范围内灰度-厚度的线性关系。通过对单一材质和多材质物体透照图像的融合试验表明：本文方法对不同材质、厚度分布的大厚度比物体均具有较强的适用性。针对融合后的高动态范围图像与低动态范围显示设备间的矛盾问题，本文研究了高动态范围图像的灰度压缩算法，利用对数域梯度信息的非线性加权处理，在增强较小结构信息梯度的同时压缩较大梯度信息，进而通过对梯度信息进行重建获取灰度图像，实验表明经过本文方法压缩增强后的图像不同厚度区域的结构信息均能够同时在低动态范围显示设备上进行完整显示。更多还原

【Abstract】 When one object with large thickness ratio is detected by the X-ray digital imagingtechnology, the traditional imaging mode with fixed tube voltage cannot achieve the completerepresentation of the object’s structure because the phenomenon of overexposure and oweexposure appears in one image, which will lead to the serious lack of the information. Atpresent, people usually adopt some detection methods to resolve this problem, such asthickness or density compensation, local irradiation, imaging device with large chargecapacity or high dynamic range. Although the above methods increase the transmission doseand the costs but is difficult to get the complete visual effect of the object.This paper will study the high dynamic range X-ray imaging technology through thedesign of system control, data acquisition, image fusion, and visualization. The main purposeis to achieve the goal of extending the dynamic range of imaging system. By selecting thethickness segmentation degree of the object and linear scale of gray values from object withsimilar thickness, we used X-ray imaging system with low dynamic range to realize imagingof several object areas with different thickness. Then the features of imaging data fromdifferent object areas were extracted and fused so that we can obtain the X-ray image withhigh dynamic range of the whole object. In order to display the fused X-ray image in thedevice with low dynamic range, we studies gray-scale compression technology of the imagewith high dynamic range. It can compress the dynamic range of the fused image, but enhanceits structure information.The article puts forward to use a high pressure X-ray source with a lineargradient voltage. The voltage can realize a linear gradient adjustment according to the setof kV Tcurve cooperating with the adjustment of exposure time interval of the detector sothat we can acquire X-ray image sequences with different tube voltage intervals rapidly. Bystudying rays with the effect of the object and gray level change rule in X-ray detection process, the thickness repetition rate under adjacent tube voltages, linear proportionalrelationship between different image gray scales and correlation characteristics of tubevoltage intervals are determined. A calculating method based on proportional gain histogramis presented to finish fast calculation of proportional gain coefficient, linear degree and repeatrate between adjacent tube voltages. Also the needed constraint of images acquired underadjacent tube voltage to the thickness repeat rate of the object is raised. It can realizeintelligent control of the second sampling process of original images and extraction of theuseful information of images from areas with different thickness. According to theproportional gain histogram, the scaling values of the needed tube voltage in different areascan be calculated. Based on it, the grey value of each pixel of the image were weighted toobtained a fused image with high dynamic range. Such fusion method can ensure linearrelationship between gray values and the thickness of the object within the scope of certainthickness changes. The fusion experiments results of X-ray images of objects made of asingle or more than one material shows that the presented method is suitable for X-raydetection of objects with different material, thickness distribution and thickness ratio.In order to display the fused image with high dynamic range on device with lowdynamic range, the gray-scale compression algorithm is studied. The image is processed bylogarithmic operation, gradient operation and nonlinear weighted operation. It can compressthe large gradient information, but enhance small gradient information. Then the gradientinformation are reconstructed to obtain gray images. The experiment results show that thecompressed image contains information of different areas of object with different thicknessand can be displayed in device with low dynamic range.更多还原