Using Graphics Processing Units to Parallelize the FDK Algorithm for Tomographic Image Reconstruction

The paper presents the implementation of a parallel version of FDK （Felkamp, David e Kress） algorithm using graphics processing units. Discussion was briefly some elements the computed tomographic scan and FDK algorithm; and some ideas about GPUs （Graphics Processing Units） and its use in general purpose computing were presented. The paper shows a computational implementation of FDK algorithm and the process of parallelization of this implementation. Compare the parallel version of the algorithm with the sequential version, used speedup as a performance metric. To evaluate the performance of parallel version, two GPUs, GeForce 9400GT （16 cores） a low capacity GPU and Quadro 2000 （192 cores） a medium capacity GPU was reached speedup of 3.37.

Reduction of artifacts in dental cone beam CT images to improve the three dimensional image reconstruction

Cone-beam CT (CBCT) scanners are based on volumetric tomography, using a 2D extended digital array providing an area detector [1,2]. Compared to traditional CT, CBCT has many advantages, such as less X-ray beam limitation, and rapid scan time, etc. However, in CBCT images the x-ray beam has lower mean kilovolt (peak) energy, so the metal artifact is more pronounced on. The position of the shadowed region in other views can be tracked by projecting the 3D coordinates of the object. Automatic image segmentation was used to replace the pixels inside the metal object with the boundary pixels. The modified projection data, using synthetically Radon Transformation, were then used to reconstruct a new back projected CBCT image. In this paper, we present a method, based on the morphological, area and pixel operators, which we applied on the Radon transformed image, to reduce the metal artifacts in CBCT, then we built the Radon back project images using the radon invers transformation. The artifacts effects on the 3d-reconstruction is that, the soft tissues appears as bones or teeth. For the preprocessing of the CBCT images, two methods are used to recognize the noisy black areas that the first depends on thresholding and closing algorithm, and the second depends on tracing boundaries after using thresholding algorithm too. The intensity of these areas is the lowest in the image than other tissues, so we profit this property to detect the edges of these areas. These two methods are applied on phantom and patient image data. It deals with reconstructed CBCT dicom images and can effectively reduce such metal artifacts. Due to the data of the constructed images are corrupted by these metal artifacts, qualitative and quantitative analysis of CBCT images is very essential.

Image processing based three-dimensional model reconstruction for cross-platform numerical simulation

Numerical simulation is the most powerful computational and analysis tool for a large variety of engineering and physical problems.For a complex problem relating to multi-field,multi-process and multi-scale,different computing tools have to be developed so as to solve particular fields at different scales and for different processes.Therefore,the integration of different types of software is inevitable.However,it is difficult to perform the transfer of the meshes and simulated results among software packages because of the lack of shared data formats or encrypted data formats.An image processing based method for three-dimensional model reconstruction for numerical simulation was proposed,which presents a solution to the integration problem by a series of slice or projection images obtained by the post-processing modules of the numerical simulation software.By means of mapping image pixels to meshes of either finite difference or finite element models,the geometry contour can be extracted to export the stereolithography model.The values of results,represented by color,can be deduced and assigned to the meshes.All the models with data can be directly or indirectly integrated into other software as a continued or new numerical simulation.The three-dimensional reconstruction method has been validated in numerical simulation of castings and case studies were provided in this study.

Underwater image clarifying based on human visual colour constancy using double-opponency

Underwater images are often with biased colours and reduced contrast because of the absorption and scattering effects when light propagates in water.Such images with degradation cannot meet the needs of underwater operations.The main problem in classic underwater image restoration or enhancement methods is that they consume long calcu-lation time,and often,the colour or contrast of the result images is still unsatisfied.Instead of using the complicated physical model of underwater imaging degradation,we propose a new method to deal with underwater images by imitating the colour constancy mechanism of human vision using double-opponency.Firstly,the original image is converted to the LMS space.Then the signals are linearly combined,and Gaussian convolutions are per-formed to imitate the function of receptive fields(RFs).Next,two RFs with different sizes work together to constitute the double-opponency response.Finally,the underwater light is estimated to correct the colours in the image.Further contrast stretching on the luminance is optional.Experiments show that the proposed method can obtain clarified underwater images with higher quality than before,and it spends significantly less time cost compared to other previously published typical methods.

Multi-scale cross-domain alignment for person image generation

Person image generation aims to generate images that maintain the original human appearance in different target poses.Recent works have revealed that the critical element in achieving this task is the alignment of appearance domain and pose domain.Previous alignment methods,such as appearance flow warping,correspondence learning and cross attention,often encounter challenges when it comes to producing fine texture details.These approaches suffer from limitations in accurately estimating appearance flows due to the lack of global receptive field.Alternatively,they can only perform cross-domain alignment on high-level feature maps with small spatial dimensions since the computational complexity increases quadratically with larger feature sizes.In this article,the significance of multi-scale alignment,in both low-level and high-level domains,for ensuring reliable cross-domain alignment of appearance and pose is demonstrated.To this end,a novel and effective method,named Multi-scale Crossdomain Alignment(MCA)is proposed.Firstly,MCA adopts global context aggregation transformer to model multi-scale interaction between pose and appearance inputs,which employs pair-wise window-based cross attention.Furthermore,leveraging the integrated global source information for each target position,MCA applies flexible flow prediction head and point correlation to effectively conduct warping and fusing for final transformed person image generation.Our proposed MCA achieves superior performance on two popular datasets than other methods,which verifies the effectiveness of our approach.

Capsule networks embedded with prior known support information for image reconstruction

Compressed sensing(CS)has been successfully applied to realize image reconstruction.Neural networks have been introduced to the CS of images to exploit the prior known support information,which can improve the reconstruction quality.Capsule Network(Caps Net)is the latest achievement in neural networks,and can well represent the instantiation parameters of a specific type of entity or part of an object.This study aims to propose a Caps Net with a novel dynamic routing to embed the information within the CS framework.The output of the network represents the probability that the index of the nonzero entry exists on the support of the signal of interest.To lead the dynamic routing to the most likely index,a group of prediction vectors is designed determined by the information.Furthermore,the results of experiments on imaging signals are taken for a comparation of the performances among different algorithms.It is concluded that the proposed capsule network(Caps Net)creates higher reconstruction quality at nearly the same time with traditional Caps Net.

Three-Dimensional Model Reconstruction of Nonwovens from Multi-Focus Images

The three-dimensional(3D)model is of great significance to analyze the performance of nonwovens.However,the existing modelling methods could not reconstruct the 3D structure of nonwovens at low cost.A new method based on deep learning was proposed to reconstruct 3D models of nonwovens from multi-focus images.A convolutional neural network was trained to extract clear fibers from sequence images.Image processing algorithms were used to obtain the radius,the central axis,and depth information of fibers from the extraction results.Based on this information,3D models were built in 3D space.Furthermore,self-developed algorithms optimized the central axis and depth of fibers,which made fibers more realistic and continuous.The method with lower cost could reconstruct 3D models of nonwovens conveniently.

Terrain reconstruction from Chang'e-3 PCAM images

The existing terrain models that describe the local lunar surface have limited resolution and accuracy, which can hardly meet the needs of rover navigation,positioning and geological analysis. China launched the lunar probe Chang＇e-3 in December, 2013. Chang＇e-3 encompassed a lander and a lunar rover called ＂Yutu＂（Jade Rabbit）. A set of panoramic cameras were installed on the rover mast. After acquiring panoramic images of four sites that were explored, the terrain models of the local lunar surface with resolution of 0.02 m were reconstructed. Compared with other data sources, the models derived from Chang＇e-3 data were clear and accurate enough that they could be used to plan the route of Yutu.

A fast algorithm for image reconstruction based on sparse decomposition

It is very slow at present to reconstruct an image from its sparse decomposition results.To overcome this one of the main drawbacks in image sparse decomposition,the property of the energy distribution of atoms is studied in this paper.Based on the property that energy of most atoms is highly concentrated,an algorithm is proposed to fast reconstruct an image from atoms’parameters by limiting atom reconstruction calculating within the atom energy concentrating area.Moreover,methods for fast calculating atom energy and normalization are also put forward.The fast algorithm presented in this paper improves the speed of the image reconstructing by approximately 32 times without degrading the reconstructed image quality.

FITTING CORRECTION METHOD OF RING ARTIFACTS FOR RECONSTRUCTING CONE-BEAM CT IMAGES

In high-resolution cone-beam computed tomography （CBCT） using the flat-panel detector, imperfect or defect detector elements cause ring artifacts due to the none-uniformity of their X-ray response. They often disturb the image quality. A dedicated fitting correction method for high-resolution micro-CT is presented. The method converts each elementary X-ray response curve to an average one, and eliminates response inconsistency among pixels. Other factors of the method are discussed, such as the correction factor variability by different sampling frames and nonlinear factors over the whole spectrum. Results show that the noise and artifacts are both reduced in reconstructed images

70kVp联合深度学习图像重建算法对胸主动脉血管内修复术后双低CTA图像质量的影响

目的探讨70 kVp联合深度学习图像重建(DLIR)算法在低辐射剂量低对比剂用量的情况下对胸主动脉血管内修复术(TEVAR)后主动脉CT血管成像(CTA)图像质量的影响。方法前瞻性纳入65例TEVAR术后需接受主动脉CTA扫描的患者,随机分为两组。常规剂量组(A组)接受100 kVp管电压和60 mL对比剂扫描,低剂量组(B组)接受“双低”方案:管电压70 kVp,对比剂用量为0.5 mL·kg^(-1)。A组图像采用50%多模型迭代重建算法(ASIR-V)重建,B组图像分别采用滤波反投影(FBP)算法(B1亚组)、50%ASIR-V(B2亚组)、90%ASIR-V(B3亚组)及DLIR-H算法(B4亚组)重建。测量并比较各组主动脉CT值、图像噪声、对比噪声比(CNR)及品质因数(FOM)。采用5分制对图像噪声、血管锐利度和整体图像质量进行评估。结果B4亚组噪声最低,较A组降低约23.49%(P<0.001)。A组、B4亚组主动脉各感兴趣区(ROI)CT值、CNR值及图像主观评分差异无统计学意义(P>0.05);B4亚组FOM值大于A组(P<0.001)。与A组比较,B组辐射剂量降低约52.38%(P<0.001),对比剂用量和注射流率分别降低约39%和33.6%(P<0.001)。结论采用70 kVp结合DLIR算法在TEVAR术后主动脉CTA扫描中可获得与常规剂量相当的图像质量,且辐射剂量和对比剂用量大幅度降低。

基于自适应深度先验的高光谱图像超分辨率

为了解决现有的高光谱超分辨率方法依赖于手工先验和数据驱动先验会导致参数选择困难或可解释性差的问题,采用一种基于自适应深度先验正则的高光谱图像超分辨率方法,进行了理论分析和实验验证。首先设计基于卷积神经网络的多阶段特征提取网络,提取退化图像的空间和光谱信息;其次将提取到的空-谱先验输入基于transformer模型的特征融合模块;然后自适应交互空域和谱域的互补信息,以捕获图像的全局先验特征;最后在退化模型中插入深度先验正则项,将超分辨率问题表述为一个优化问题,其解可以通过交替方向乘子法获得并降低求解复杂度。结果表明,所提出算法在信噪比均为35 dB时,重建信噪比分别达到了34.16 dB和29.35 dB,比次优算法高出2.78 dB和2.17 dB,重建的高分辨率高光谱图像与其固有结构具有较高的一致性。该研究为综合利用手工先验和数据驱动先验增强高光谱图像空间分辨率提供了参考。

数字图像处理在结构优化结果提取中的应用

结构优化结果的提取和应用主要考虑如何有效地将优化结果转化为可用的CAD模型,实现CAE和CAD之间的数据共享和模型转换。基于数字图像原理提取轮廓范围及像素点坐标,根据相邻点坐标计算离散曲率,并根据曲率值对构件的边缘形状进行分类,得到拓扑优化结果。该方法提高了拓扑优化结果的精度和可用性及提取的效率,减少了人为因素的干扰,提取的模型更符合实际,有助于更好地理解和分析拓扑优化结果,对拓扑优化后处理的结果提取轮廓重构及成果应用有一定参考作用。

一种利用光子图像解决深度学习中子CT缺乏成对数据的方法

由于缺少高质量成对数据集,深度学习在中子层析(CT)重建中的应用与发展受到了严重阻碍。虽然中子层析和光子层析成像原理都基于Radon变换,但二者在粒子输运过程中的成像特征不同,所以光子层析图像训练的网络无法直接用于解决中子层析成像的重建问题。为此,本文提出了一种可以解决光子和中子层析成像迁移过程中概率分布差异问题的无监督域适应网络。该方法通过引入最大均值差异计算以减小光子与中子层析图像特征之间的分布差异,并通过小波变换与卷积神经网络相结合的方式增强重建的有效特征。与其他算法的对比验证表明,该方法能够从低投影通量下的中子层析结果中重建出高质量的中子层析图像,有效缓解了低投影通量下中子层析成像退化的问题。

CT图像后处理方法对颅内动脉瘤形态学参数测量准确性的影响

目的:以三维数字减影血管造影(three-dimensional digital subtraction angiography,3D-DSA)为“金标准”,基于CT血管成像(CT angiography,CTA)对比最大密度投影(maximum intensity projection,MIP)联合多平面重组(multiplanar reconstruction,MPR)和减影的容积再现(volume rendering,VR)这2种图像后处理方法之间颅内动脉瘤形态学参数测量结果的差异。方法:回顾性分析重庆医科大学附属第一医院92例(共116个颅内未破裂动脉瘤)患者的头颈CTA和DSA影像资料。使用双盲法分别在3D-DSA的VR、CTA的MIP-MPR和减影VR图像上测量动脉瘤的形态学参数,测量2次取均值,并计算相关的形态学衍生指标。使用组内相关系数(intraclass correlation coefficient,ICC)比较观察者间测量值的一致性,利用重复测量的方差分析或Friedman秩和检验比较不同后处理图像上测量结果的差异,后续组间两两比较采用Bonferroni校正。采用卡方检验对比各组小、微小动脉瘤及宽颈动脉瘤的分布情况。结果:基于MIP-MPR、减影VR和3D-DSA测得的动脉瘤最大径、瘤宽和瘤高在三组间整体比较时差异无统计学意义(χ^(2)值分别为4.484、5.808、4.468,均P>0.05);而得到的颈宽、载瘤动脉的平均直径在三组间整体比较时差异有统计学意义(χ^(2)值分别为48.341、87.948,均P<0.001);进行后续的组间两两比较时结果存在差异,且这种差异在颈内动脉虹吸部与非虹吸部2个亚组内又有一定区别。由MIP-MPR得出的形态学衍生指标均偏大,与DSA相比差异均有统计学意义(P<0.017),除高宽比外,与减影VR相比差异也有统计学意义(P<0.017)。基于减影VR进行动脉瘤的形态学评估,仅载瘤动脉的平均直径和大小比与DSA相比差异有统计学意义(P<0.017),但载瘤动脉的平均直径测量差异在亚组分析时不明显(P>0.017);其余参数与DSA相比差异均无统计学意义(P>0.017)。结论:减影VR上颅内动脉瘤形态学参数测量的准确性优于MIP-MPR,更适用于颈内动脉虹吸部动脉瘤的形态学评估。

极化雷达图像目标超分辨率重建研究进展

成像雷达具有全天时、全天候的观测能力,能够通过成像处理获得目标雷达图像信息,是对地观测、侦察监视等民用和军用领域中的重要遥感设备。高分辨率雷达图像能够提供目标的详细轮廓和精细结构,有利于后续目标分类识别等应用。对获取的雷达图像,如何利用信号和信息处理等理论方法进一步提升分辨率,突破分辨率瑞利极限,具有重要的科学研究和实际应用价值。另一方面,作为电磁波的重要属性之一,极化在目标特性的获取和挖掘中发挥着重要作用,能够为目标超分辨率重建带来丰富信息。为此,该文梳理了极化雷达图像目标超分辨率重建的概念及性能评价指标,并重点归纳整理了极化雷达图像目标超分辨率重建方法及其应用。最后,总结了现有方法的局限性并展望了未来的技术发展趋势。

基于线阵相机的水下射弹动态参数及超空泡演化过程测量方法

针对水下环境恶劣、光线条件差、难以拍摄并复原高速射弹运动姿态及其超空泡演化过程等问题,采用线阵相机的交汇测量原理,设计基于线阵相机的水下射弹图像采集系统、搭建基于双线阵相机的图像采集靶面,提出一种针对水下高速超空泡目标识别与弹形复原算法。利用超空泡射弹几何关系与相机标定数据推导出射弹过靶坐标、姿态角与过靶速度公式,搭建12.7 mm滑膛枪垂直发射入水实验,对系统的可行性进行验证分析。实验结果表明:该系统能够有效地进行水下射弹图像的快速采集,较完整的对弹形及空泡形态进行分离与复原,解算出射弹水下运动速度、姿态角及超空泡形态参数;系统解算出的速度与同时部署的高速摄像所处理的速度误差在1%以内,证明该系统具有较高的可靠性与实用性。

基于C-MC算法的头部CT图像三维重建

为了制作适用于伽马刀手术的3D打印医疗头模,基于断层CT扫描图像,首次提出Canny和MC相结合的算法(C-MC)进行人体头部的三维重建。Canny边缘检测算法采用高斯滤波去噪,以便于加强边缘和边缘提取,与传统移动立方体(Marching Cubes)算法相结合,完成头部图像的快速、精确三维重建与可视化。使用C-MC算法得到的人体头部三维重建模型,不仅提高了三维重建的效率和头部图像的完整性,也为后期医疗头模的三维建模设计以及3D打印提供了有力的支撑依据。

基于主动位移成像的图像超分辨率重建

超分辨重建算法是一种将低分辨率图像恢复为高分辨率图像的算法,被广泛用于医学、遥感、军事安防以及人脸识别等领域。在黑夜、远场场景下构建数据集比较困难,基于深度学习的超分辨重建算法应用受到阻碍。而微扫描成像技术扫描模式固定,对器件到位精度要求高。针对这两个问题,我们提出一种基于主动位移成像的图像超分辨率重建算法。具体地,在控制相机随机移动的同时记录采样时刻位移,通过解算、映射选图、精确匹配图像序列并获取多帧图像间的亚像素信息,然后对估计图像进行迭代和更新,最后重建获得高分辨率图像。实验结果表明,本算法在PSNR、SSIM和平均梯度三个指标上都优于最近提出的基于POCS的图像超分辨率重建算法MFPOCS,与基于CNN的方法ACNet相比具有竞争力。值得提出的是,本算法无需固定的扫描模式,降低了微扫描技术对器件实时到位精度的要求,同时,本算法可以保证重建初始帧的优良选取,有效规避了POCS算法的固有缺点。

基于双目立体视觉的实验室波浪场实时重建

准确测量波浪场演化对于海洋结构物的水动力研究十分重要。本文基于双目立体视觉建立了一套用于测量室内波浪的图像识别方法,以轻质泡沫作为示踪粒子标记波面以获得具有丰富纹理的波面图像,根据极线约束矫正图像,使用阈值分割避免光照不均和水底反射的干扰,使用GPU加速的立体视觉匹配算法,实现波浪图像的实时立体匹配,重建出波面的三维点云与网格模型。通过对图像序列的重建,提取出的波浪时历与浪高仪的测量结果基本一致,波高以及周期值的吻合性良好。结果表明,该方法可以准确测量实验室内的三维波浪场,捕捉瞬时波浪演化,并可提供实时的监控。