本文關鍵詞：角度受限下納米CT圖像處理及三維重構算法研究　出處：《中國科學技術大學》2016年博士論文　論文類型：學位論文

  更多相關文章： 軟X射線納米CT 硬X射線納米CT 投影角度受限 圖像預處理 三維重構算法 全變分 多灰度圖像

【摘要】：近些年,隨著X射線光源性能的不斷進步,高靈敏度、高穩(wěn)定性的X射線探測器的開發(fā),微納米級X射線光學元件加工技術的提升,精密機械平臺及控制系統(tǒng)的使用,基于波帶片的X射線顯微成像結合傳統(tǒng)CT方法得到的納米CT成像技術得到顯著地發(fā)展。其中,硬X射線納米CT具有強穿透能力、大焦深、高分辨、無需真空環(huán)境、包含多種材料吸收邊、三維成像等優(yōu)勢,已經在材料科學、微電子產業(yè)、環(huán)境工程、能源科學等眾多領域得到廣泛應用。軟X射線納米CT對于生物樣品具有天然的高襯度、能夠保持樣品“鮮活”時的結構和形態(tài)等特性,已經應用于藥物研發(fā)、細胞機理、納米顆粒等諸多領域。因此納米CT成像技術具有其獨特優(yōu)勢,能夠彌補電鏡、光學顯微鏡等其它顯微成像手段的一些不足。納米CT系統(tǒng)采集的投影數(shù)據(jù)需要經過數(shù)據(jù)預處理、三維重構、圖像分割、后處理分析計算等一系列的處理步驟,最終才能得到準確、清晰、有意義的信息。投影圖像的預處理是后續(xù)系列處理步驟的前期準備,需要校正采集的數(shù)據(jù)中存在的壞點、亮度不均勻、投影間光強不相等、噪聲、轉軸偏差等。而三維重構是納米CT應用過程的最關鍵步驟,是把疊加的信息分離的過程,良好的圖像重構質量能真實反映樣品三維結構信息,也是進行后處理計算的前提。然而,三維重建常常面臨投影數(shù)據(jù)不完備的難題,在納米CT數(shù)據(jù)重建中,經常遇到投影角度受限的情況,因此迫切需要發(fā)展合適的重構算法加以解決。國家同步輻射實驗室于2007年左右建成了一臺硬X射線納米CT設備,隨后又建設了軟X射線納米CT設備,本論文結合我們平臺的數(shù)據(jù)特點,開展了以下幾方面的工作：1.總結了X射線納米顯微成像技術的優(yōu)勢和特性,并分別綜述了硬X射線納米CT和軟X射線納米CT的工作原理及主要光學元件的原理與參數(shù)。我們選擇了幾個典型例子來說明硬X射線納米CT具有的獨特優(yōu)勢：對硅酸鋅納米棒自組裝催化劑顆粒成像,內部結構與缺陷能夠清晰呈現(xiàn),這是電鏡等其它成像手段所不能的：對鋰離子電池中的氧化銅電極進行鋰化-反鋰化過程原位動態(tài)成像,氧化銅顆粒等隨電池充放電過程的變化被清晰記錄：對有絲分裂的酵母細胞成像,體現(xiàn)了相襯成像模式下,硬X射線納米CT對生物大樣品能成高襯度像的能力。為了展示軟X射線納米CT的應用價值,我們分別選取了牛痘病毒感染細胞、瘧原蟲感染紅細胞、藥物對細胞的作用、免疫金標記的干細胞、微凝膠顆粒與油滴顆粒相互作用等進行成像的例子,同時還列舉了其與熒光顯微鏡聯(lián)合成像,以及應用于近邊X射線吸收光譜成像等來體現(xiàn)軟X射線納米CT目前的應用領域。2. 發(fā)展了多種納米CT圖像預處理算法。對投影數(shù)據(jù)進行一系列的預處理,提出了基于金顆粒點軌跡追蹤、中心骨架提取,并進行中心骨架的正弦曲線預估與位移校正,來達到轉軸自動校正的目標：根據(jù)投影數(shù)據(jù)特點,在單幅投影光強均衡化的基礎上,提出了不同投影間光強均衡的方法,使得所有投影更加均勻：針對投影中空白部分(背景)進行閾值濾波,減少誤差,達到背景“干凈”的目標,為三維重構、后期分析處理等打下良好基礎。3.研究CT成像的基本原理,分析比較了傳統(tǒng)濾波反投影算法和迭代重構算法的特點及不足；并分析了納米CT成像數(shù)據(jù)特點,研究了能突破奈奎斯特采樣定理的壓縮感知理論。在此基礎上,我們改進并發(fā)展了適用于納米CT數(shù)據(jù)的基于全變分的納米CT重構算法。并通過一系列模擬數(shù)據(jù)來對比在投影中含有噪聲,以及投影角度稀疏、投影角度受限等數(shù)據(jù)不完備情況下,算法與傳統(tǒng)重構算法的重構效果好壞；還對硬X射線納米CT上采集的投影角度受限的酵母細胞數(shù)據(jù),及從軟X射線納米CT上采集的投影角度受限的碳納米管數(shù)據(jù)分別采用基于全變分的納米CT重構算法與傳統(tǒng)重構算法進行了重構并對比分析,發(fā)現(xiàn)前者優(yōu)勢明顯。4.針對納米CT、工業(yè)CT、電鏡等斷層成像中經常遇到的離散樣品重構問題,本文在充分研究分析了現(xiàn)有的離散斷層重構算法的基礎上,著力于解決現(xiàn)有算法對樣品灰度先驗知識的過分依賴,以及當圖像灰度數(shù)量增多時,重構質量下降明顯的問題。本文提出了一種以每個單獨區(qū)域作為一個研究對象的多灰度圖像離散斷層重構算法,算法繼承了離散斷層重構算法對投影角度受限強耐受力等優(yōu)點的同時,還避免了對圖像灰度先驗知識的依賴等劣勢。通過一系列實驗驗證了算法的穩(wěn)定性、收斂性、重構精度、灰度極限數(shù)量等各項性能指標,從而證明了此算法能夠突破多組分離散樣品在投影角度受限情況下定量重構分析這一瓶頸,為有效提升離散斷層成像方法在納米CT等斷層成像中的廣泛應用提供了更廣大空間。
[Abstract]:In recent years, with the continuous progress of X ray source performance, the development of X ray detector with high sensitivity and high stability, the micro nano X X-ray optics processing technology, precision mechanical platform and control system, nano CT imaging imaging system X ray microscope combined with traditional CT methods were statistically significant based on the development of. Among them, the hard X - ray CT nanoparticles have strong penetrating ability, large focal depth and high resolution, no vacuum environment, including edge, three-dimensional imaging and other advantages of various materials absorption, has been widely used in many fields of material science, microelectronics industry, environmental engineering and energy science. Soft X ray nano CT has been applied to many fields such as drug discovery, cell mechanism, nanoparticle and so on. Therefore, the nano CT imaging technology has its unique advantages, which can make up for some shortcomings of other microscopic imaging methods, such as electron microscope and optical microscope. The projection data collected by nano CT system need a series of processing steps, such as data preprocessing, 3D reconstruction, image segmentation, post-processing, analysis and calculation, so as to get accurate, clear and meaningful information. Preprocessing of projection image is prepare for the later series of processing steps, need to point, brightness correction data are not uniform, not equal intensity projection, noise, axis deviation. Three dimensional reconstruction is the most critical step in the application process of nano CT. It is the process of separating the superimposed information. The quality of good image reconstruction can truly reflect the three-dimensional structure information of the sample, and it is also the premise for post processing calculation. However, 3D reconstruction often faces the problem of incomplete projection data. In the reconstruction of nano CT data, projection angle is often limited. Therefore, it is urgent to develop appropriate reconstruction algorithm to solve it. In 2007, built around a hard X ray nano CT equipment in National Synchrotron Radiation Laboratory, then the construction of the soft X ray nano CT devices, this paper combined with the characteristics of our data platform, to carry out the work in the following aspects: 1. summarizes the advantages and characteristics of nano X ray imaging technology, and the principle and the parameters of the working principle of hard X ray nano CT and nano CT soft X ray and the main optical elements were reviewed. We chose some typical examples to illustrate the advantages of hard X ray nano CT: zinc silicate nanorods self-assembled catalyst particle imaging, the internal structure and defects can be clearly presented, this is the electron microscope and other imaging method can not do: lithium lithium - process in situ dynamic imaging of copper oxide lithium electrode ion batteries, copper oxide particles change with the battery charge and discharge process is clearly recorded: on yeast cell mitosis of imaging, phase contrast imaging mode, the hard X - ray CT nanoparticles on biological samples into a high degree as the ability of lining. In order to show the application value of soft X ray nano CT, we selected the vaccinia virus infected cells and Plasmodium falciparum infected red blood cells and drugs on cell function, immunogold labeling of stem cells, the microgel particles and oil droplets interactions such as imaging examples, and also lists the combined imaging and fluorescence microscope, and applied to the near edge X-ray absorption spectroscopy X imaging to reflect the application of soft X ray nano CT current. 2. a variety of nano CT image preprocessing algorithms have been developed. The projection data preprocessing is proposed to extract a series of gold particles trajectory point tracking and framework based on the sine curve and central frame prediction and displacement correction, to achieve the goal: shaft automatic correction according to the characteristics of the projection data, based on a single projection intensity equalization, and puts forward the method of intensity balance different projection, the more uniform for all projection projection blank part (background) threshold filtering, reduce errors, achieve the background of "clean" target, and lay a good foundation for the later analysis, 3D reconstruction. 3., we studied the basic principles of CT imaging, analyzed and compared the characteristics and shortcomings of traditional filtered back projection algorithm and iterative reconstruction algorithm, analyzed the characteristics of nano CT imaging data, and studied the compressive sensing theory that could break through the Nyquist sampling theorem. On this basis, we have improved and developed a fully variational based nano CT reconstruction algorithm for nanoscale CT data. And through a series of simulation data to compare the noise in the projection, and the projection angle, sparse projection angle limited data in case of incomplete, or reconstruction algorithm and traditional reconstruction algorithm; yeast cell data of hard X ray projection angle acquisition on nano CT limited, the data collected from the soft and carbon nanotubes X ray nano CT projection of the limited angle respectively by nano CT total variation reconstruction algorithm and traditional reconstruction algorithm for the reconstruction and comparative analysis based on the former found obvious advantages. 4. aiming at the problem of discrete sample reconstruction often encountered in nano CT, industrial CT, electron microscope tomography, this paper analyzes on the basic of research on discrete tomography reconstruction algorithm on the current focus on solving the existing algorithms over reliance on sample gray prior knowledge, as well as the increase in the number of gray image, the obvious problem of reconstruction quality drop. This paper puts forward a method to separate each region as a research object of the discrete fault reconstruction algorithm of grayscale image, the algorithm inherits the discrete fault reconstruction algorithm of limited projection angle strong tolerance advantages but also avoids the disadvantage of image prior knowledge dependence. Through a series of experiments, the stability, convergence, precision of reconfiguration, the number of gray limit and so on are verified.
【學位授予單位】：中國科學技術大學
【學位級別】：博士
【學位授予年份】：2016
【分類號】：TP391.41

【相似文獻】

相關期刊論文 前10條

1 劉盾;石和平;;基于一種改進的壓縮感知重構算法的分析與比較[J];科學技術與工程;2012年21期

2 蔣英春;;離散空間中正交小波分解重構算法的實現(xiàn)[J];計算機應用研究;2013年02期

3 劉勇;魏東紅;毛京麗;;基于優(yōu)化內積模型的壓縮感知快速重構算法[J];北京郵電大學學報;2013年01期

4 王田川;宋建新;;壓縮感知重構算法研究[J];電視技術;2013年11期

5 李福建,陳廷槐,田梅,周六丁;一種新的環(huán)網(wǎng)故障診斷與重構算法[J];計算機工程;1992年06期

6 童露霞;王嘉;;基于壓縮傳感的重構算法研究[J];電視技術;2012年11期

7 李博;郭樹旭;;一種改進的壓縮感知重構算法研究[J];現(xiàn)代電子技術;2013年03期

8 李志剛;;一種快速的壓縮感知信號重構算法[J];信息技術;2013年06期

9 梁棟,楊尚俊,章權兵;一種基于圖象序列的3D重構算法[J];安徽大學學報(自然科學版);2001年01期

10 陳勤;鄒志兵;張e，

本文編號：1345015