平板探测器的图像性能优化(硕士)(论文30000字)
THE IMAGE PERFORMANCE OPTIMIZATION OF FLAT PANEL DETECTOR
摘 要
X射线成像技术在医疗、工业探伤、航空航天等众多领域得到广泛的应用。平板探测器因为拥有品质突出、数据传输便利、性能稳定等优势而深受青睐。
集成电路的规模化发展以及非晶硅产业的突飞猛进，推动着数字X-Ray平板探测器行业的发展高潮。随着人类数字化步伐的加快，X-Ray成像技术借着数字化信息发展的东风迈着大步向数字信息时代跨进。随着科学技术的发展，平板探测器(FPD)作为数字成像革命的关键性产品进入人们的视野，人们将之称为影像增强仪后常规X-Ray射线成像领域的最大一次革命[1]。当前，医疗体外诊断、工业无损探伤以及PCBA生产检查是是X射线数字成像技术的主要应用领域，但在其他领域的应用也具有广阔的发展前景。
本文为大家详细介绍了DR成像系统的发展历程以及其各个部分的功能。探测器的发展经历了胶片成像，计算机成像直至如今的数字成像时代，随着技术的突破，DR图像已经具备快捷、稳定以及图像性能突出等优点。平板探测器主要包含转换介质、图像采集单元以及图像传输单元。平板探测器最显著的特点是体积小，质量轻，便于携带且图像性能突出。受限于各种各样的原因，成像过程中必然会引入噪声使图像质量降低，存在影响医生的诊断正确性的风险。因此，定位影响图像质量的噪声来源、噪声特性以及信噪之间的关系，并对图像信号作降噪处理，对图像细节作增强处理，提升图像的质量成为平板探测器研究的重要方向。
本文研究了DR系统噪声的形成原因，其一般分为暗电流噪声、不均匀性以及基于康普顿效应引起的随机噪声。利用多一般的处理措施能够有效处理掉绝大部分噪声信号，但高斯噪声会极大程度的影响图像质量且处理极其困难。基于目前普遍应用的高斯模型以及拉普拉斯模型对DR图像中的多尺度高频信息无法有效的描述，噪声处理效果并不理想，本文主要应用的是基于 Laplace- Impact混合模型实现的最小均方误差估计去噪算法（MMSE）。此算法线利用双树复小波对DR图像进行分解处理，再利用局部的均方差数值对系统的噪声参数进行估计，之后通过 MMSE估计实现对高频系数的优化，最后基于逆小波变换实现其高频小波系数的优化，并重新转化为图像。根据实验结果,LI-MMSE算法在对高斯噪声的处理中明显优于BLM-GSM和SoftLMap这两种图像处理算法。
关键词：X射线，平板探测器，图像，噪声，拉普拉斯，LI-MMSE算法
THE IMAGE PERFORMANCE OPTIMIZATION OF FLAT PANEL DETECTOR
ABSTRACT
X-ray imaging technology has been widely used in many fields such as medical treatment, industrial exploration and aerospace. Because of its advantages such as high quality, convenient data transmission and stable performance, the flat detector is very popular.
The scale development of integrated circuits and the advance of amorphous silicon industry are driving the development of the X-ray flat panel detector. With the acceleration of the digitization of human beings, X-ray imaging technology is stepping into the digital information age with the development of digital information. With the development of science and technology, flat panel detector (FPD) as a key product of digital imaging revolution into people's field of vision, people called it as the biggest revolution in the field of conventional X-Ray imaging after Image intensifier. At present, medical diagnosis, industrial nondestructive flaw detection in vitro and PCBA production inspection is the main application field for X-ray digital imaging, but it also has a board development prospects for the other application.
The document introduces the development course of X-ray imaging system and its functions. The detector experienced from the film imaging, computer imaging to digital imaging, with the technology breakthrough, DR image has the advantage of fast, stable and outstanding performance. The plate detector mainly includes conversion medium, image acquisition unit and image transmission unit. The most notable feature of the panel detector is small volume, light weight, easy to carry and prominent image performance. Due to all kinds of reasons, it is inevitable to introduce noise in the imaging process to reduce the quality of the image, and there is a risk of influencing the accuracy of the doctor's diagnosis. Therefore, affect the image quality of noise source, noise characteristic and the relation between the signal-to-noise, the image signal and the noise reduction processing, enhancement of image detail processing, improve the quality of the image has become an important direction of flat panel detector research.
The document researches the formation cause of the imaging noise of DR system, which is mainly divided into the dark current noise, the inhomogeneity and the random noise caused by Compton effect. Using multiple frames to calculate the average, multipoint linear fitting, and average filtering treatment measures can effectively get rid of most of the noise signal, but the Gaussian noise would greatly affect image quality and deal with extremely difficult. As the generally applied Gaussian and Laplace model can’t effectively described the multi-scale high frequency information of the DR image and the noise processing effect is not ideal, the treatment measure of the document is the minimum mean square error (MMSE) which is based on the Laplace - Impact mixed model.
At first, the algorithm to use the dual tree complex wavelet to make the DR images decomposition, then through the MMSE estimation approach to achieve the optimization of high frequency coefficients, at last use the inverse wavelet transform approach to achieve the optimization of the high frequency wavelet coefficients, and to convert into images. According to the experiment results, the LI-MMSE algorithm is significantly better than BLM-GSM and SoftLMap in the processing of gaussian noise.
Keywords: X-ray, tablet detector, image, noise, Laplace, LI-MMSE algorithm
目录
摘 要 I
ABSTRACT II
第一章 绪论 1
1.1 X-Ray探测发展简介 1
1.2 胶片成像系统 1
1.3 计算机X-Ray成像系统 2
1.4 数字X-Ray成像系统 3
1.5 平板探测器的意义 3
1.6 本文的主要内容 4
第二章 X光系统简介 5
2.1 X射线成像原理 5
2.2 DR系统硬件组成 6
2.2.1 高压发生器与球管 6
2.2.2 手闸与控制盒 7
2.2.3 数字化探测器 7
2.2.4 计算机系统 9
2.3 本章小结 10
第三章 数字X-Ray探测器工作原理 11
3.1 数字平板探测器的分类 11
3.1.1 以转换层分类的探测器类型 11
3.1.2 以信号检测分类的探测器类型 12
3.2 数字X-Ray探测器工作基本原理 13
3.2.1 闪烁体 13
3.2.2 TFT面板 14
3.2.3 信号采集单元 15
3.3 本章小结 16
第四章 数字图像处理理论简介 17
4.1 图像质量的基本概念 17
4.1.1 空间域表征 17
4.1.2 频率域表征 17
4.1.3 灰阶表征 18
4.1.4 图像质量评价 18
4.2 数字图像处理的基本概念 19
4.2.1 数字图像噪声 20
4.2.2 图像降噪算法 21
4.2.3 图像增强算法 23
4.3 X射线图像的质量分析 24
4.4 本章小结 25
第五章 多尺度DR图像去噪 26
5.1 系统噪声去除 26
5.1.1 暗电流校正 26
5.1.2 图像的不均匀校正 27
5.1.3 散点噪声去噪算法 29
5.2 高斯噪声多尺度去噪理论基础 30
5.2.1 噪声的尺度衰减性 30
5.2.2 小波系数的尺度间相关性 30
5.2.3 小波系数的尺度内相关性 31
5.3 高斯噪声去噪算法 31
5.3.1 BLS-GSM去噪算法 32
5.3.2 SoftLMAP去噪算法 33
5.4 基于Laplce-Impact模型的去噪算法 34
5.4.1 DT-CWT高频系数的Laplace-Impact模型 34
5.4.2 基于Laplace-Impact混合模型的MMSE估计算法 36
5.4.3 局部方差估计 36
5.5 实验结果与分析 37
5.5.1 人工噪声去噪 37
5.5.2 实际DR图像去噪 38
5.5.3 窗口大小对去噪效果的影响 40
5.6 本章小结 41
第六章 总结与展望 42
参考文献 43 |
