加权窗口在数字图像相关技术中的应用研究

应用数字图像相关法计算变形前后图像中各点的位移量时,先通过整像素搜索法来确定各点的整像素位移,进而根据整像素搜索得到的相关系数值来计算各点的亚像素位移.提出了一种新的整像素搜索法,即采用计算窗口加权的方法对变形前后图像中的点进行搜索计算.计算结果表明,该方法是可行的,同时可以提高计算结果准确度.     

A fast digital image correlation method for deformation measurement

Fast and high-accuracy deformation analysis using digital image correlation (DIC) has been increasingly important and highly demanded in recent years. In literature, the DIC method using the Newton-Rapshon (NR) algorithm has been considered as a gold standard for accurate sub-pixel displacement tracking, as it is insensitive to the relative deformation and rotation of the target subset and thus provides highest sub-pixel registration accuracy and widest applicability. A significant drawback of conventional NR-algorithm-based DIC method, however, is its extremely huge computational expense. In this paper, a fast DIC method is proposed deformation measurement by effectively eliminating the repeating redundant calculations involved in the conventional NR-algorithm-based DIC method. Specifically, a reliability-guided displacement scanning strategy is employed to avoid time-consuming integer-pixel displacement searching for each calculation point, and a pre-computed global interpolation coefficient look-up table is utilized to entirely eliminate repetitive interpolation calculation at sub-pixel locations. With these two approaches, the proposed fast DIC method substantially increases the calculation efficiency of the traditional NR-algorithm-based DIC method. The performance of proposed fast DIC method is carefully tested on real experimental images using various calculation parameters. Results reveal that the computational speed of the present fast DIC is about 120-200 times faster than that of the traditional method, without any loss of its measurement accuracy     

基于图像内容视觉感知的图像质量客观评价方法

图像质量客观评价在图像和视频传输、编解码以及服务质量中起着非常重要的作用.然而现有的方法往往没有考虑图像内容特征及其视觉感知,使得其质量客观评价与主观感知结果存在一定的差距.基于此,本文结合图像内容的复杂性特征和人眼的掩蔽特性、对比敏感度特性以及亮度感知的非线性特性,提出了一种基于人眼对图像内容感知的图像质量客观评价方法.该方法首先结合亮度感知的非线性模型将图像进行转换,得到人眼感知强度图;再分别以人眼对比敏感度值和图像局部平均对比度值作为权重因子对强度求和,以求和的数据信息作为人眼感知图像的内容,并构建图像感知模型;最后以此模型分别模拟人眼感知参考图像和失真图像,并计算二者的强度差,以强度差为评价分数的基础构建图像质量客观评价模型.采用LIVE,TID2008和CSIQ三个数据库中的共47幅参考图像和1549幅测试图像进行仿真实验,且与SSIM,VSNR,FSIM和PSNRHVS等典型的图像质量客观评价模型进行对比分析,同时探讨影响图像质量评价的因素.结果表明:所提方法的评价分数与主观评价分数的Pearson线性相关性系数和Spearman秩相关系数值比SSIM的评价结果均有一定程度的提高,提高幅度分别平均为9.5402%和3.2852%,比PSNRHVS和VSNR提高幅度更大.综合以上表明:所提方法是一种有效可行的图像质量客观评价方法;同时,在图像质量客观评价中,考虑人眼对图像内容的感知和复杂度的分析有助于提高图像质量主客观评价的一致性,评价精度可得到进一步的提高.     

一种空域和频域相结合的运动图像亚像素配准技术

针对视频图像运动检测问题,提出了一种结合空域灰度投影和频域相位相关的亚像素图像配准方法。首先,采用灰度投影算法在空域对运动图像进行粗配准,即在图像行和列方向上计算图像灰度投影特征数据,根据灰度相关函数最小化准则,估计像素级运动量;然后,在经过粗配准的两幅图像中心选取尺寸相同的区域,进行快速傅里叶变换,在频域采用扩展的相位相关算法对图像进行精确配准。该方法利用图像的功率谱信息,减少对图像内容的依赖,运用基于最小二乘的曲面拟合法,实现亚像素图像配准参数估计,具有从粗到精的特点,有效提高了图像检测精度。文中最后对样本图像进行了图像配准对比实验,结果表明,该方法可以检测0.01 pixel的运动量,最大配准误差为0.004 8 pixel。     

High accuracy digital image correlation powered by GPU-based parallel computing

A sub-pixel digital image correlation (DIC) method with a path-independent displacement tracking strategy has been implemented on NVIDIA compute unified device architecture (CUDA) for graphics processing unit (GPU) devices. Powered by parallel computing technology, this parallel DIC (paDIC) method, combining an inverse compositional Gauss–Newton (IC-GN) algorithm for sub-pixel registration with a fast Fourier transform-based cross correlation (FFT-CC) algorithm for integer-pixel initial guess estimation, achieves a superior computation efficiency over the DIC method purely running on CPU. In the experiments using simulated and real speckle images, the paDIC reaches a computation speed of 1.66×10⁵ POI/s (points of interest per second) and 1.13×10⁵ POI/s respectively, 57–76 times faster than its sequential counterpart, without the sacrifice of accuracy and precision. To the best of our knowledge, it is the fastest computation speed of a sub-pixel DIC method reported heretofore.     

Digital system of invariant correlation to position and rotation

A new correlation digital system invariant to position and rotation is presented. This new algorithm requires low computational cost, because it uses uni-dimensional signatures (vectors). The signature of the target so like the signature of the object to be recognized in the problem image is obtained using a binary ring mask constructed based on the real positive values of the Fourier transform of the corresponding image. In this manner, each image will have one unique binary ring mask, avoiding in this form the relevant information leak. Using linear and non-linear correlations, this methodology is applied first in the identification of the alphabet letters in Arial font style and then in the classification of fossil diatoms images. Also, this system is tested using the diatom images with additive Gaussian noise. The non-linear correlation results were excellent, obtaining in this way a simple but efficient method to achieve rotation and translation invariance pattern recognition.     

基于相位相关的亚像素配准技术及其在电子稳像中的应用

介绍一种基于相位相关的图像配准方法,并将该方法应用于序列图像的运动检测,通过运动补偿,实现了序列图像的稳定输出。采用相位相关配准方法对序列图像进行亚像素精度配准,运用基于最小二乘的曲面拟合法,估计当前图像相对参考图像的亚像素级的平移量。在补偿图像运动时,为保证亚像素级的补偿精度,采用平滑算法,避免在图像补偿时出现马赛克现象。最后,对样本图像进行了亚像素级位移配准和图像补偿对比实验。实验结果表明,提出的方法可以检测到0．01pixel的运动量,最大配准误差为0．008pixel;采用亚像素级运动量补偿图像,最大误差〈0．5。     

前向像移补偿下航空测绘相机畸变的纠正方法

为了提高航空测绘相机的制图精度,研究了前向像移补偿过程中测绘相机畸变的纠正方法。根据光学像差理论对畸变的影响因素进行了分析,得到了前向像移补偿过程中影响畸变的关键参数。考察了主点、主距等参数变化对畸变的影响规律,在此基础上提出前向像移补偿下畸变的计算模型。依据提出的模型对前向像移补偿下的航空测绘相机进行了数学仿真及外场飞行实验。结果表明,提出的算法稳定、有效,可以有效地纠正前向像移补偿时航空测绘相机的畸变,平面位置中误差较传统纠正方法可提高1.38倍。     

Super-resolution reconstruction face recognition based on multi-level FFD registration

Super-resolution (SR) reconstruction is an effective method to solve the problem, that the face image resolution is too low to be recognized in video, but the non-rigid change of deformed face and expression changes greatly affect the accuracy of registration and reconstruction. To solve these problems, a method of multi-level model free form deformation (FFD) elastic registration algorithm based on B spline is proposed. It first use low-resolution FFD grid for global registration, to emphasize the contribution of edge information for registration, we introduce edge registration measure into the sum of squared difference (SSD) criterion. Then divide the global registration image and reference image into a series of corresponding sub-image pairs and calculate the correlation coefficient of each pair; at the same time, we do local registration with high-resolution FFD grid to the small value correlation coefficient sub-image pairs. In the registration process of optimization, the paper uses adaptive step length gradient descent method algorithm based on chaotic variables to improve optimization efficiency. After registration, the algorithm of project onto convex sets (POCS) is used to reconstruct SR face image through several low resolution image sequences, and then recognized these SR face images by support vector machines (SVM) classifier. Experimental results from standard video database and self-built video database show that this method can register and reconstruct face image accurately in the condition of great face deformation and expression change, while the face recognition accuracy is also improved.     

基于视差自相关特征水深反演

提出了一种基于视差自相关特征模拟水深的新方法。这种方法利用不同摄站获取的同一物体的两幅图像,在水的折射影响下,造成的视觉上的偏差计算水深,可以得到相关区域三维显示中各点的水深数值。以数码相机所拍摄两张照片为例,实现相关区域水深三维显示。结果表明此算法具有一定的可行性,可以进行更深层次的研究,并应用于浅海、河流等地的资源探测开发。     

Super-resolution PIV by recursive local-correlation

This paper presents a method of accurately determining the displacement of tracer particles between two images using recursive correlation. The local correlation value is iteratively arrived at through successive approximations of local displacement using increasingly smaller regions of determination. By starting with a large search area and iteratively narrowing the search restricting the search after each iteration based on the resulting calculation, very high-resolution PIV processing can be achieved. Spurious vectors are eliminated and accuracy and processing speed is maintained by correlating images in compressed (sparse array) format using second-order spatial correlation. The methodology of this unique image analysis method is presented along with a discussion of its limitations and applicability.     

Image quality assessment using the singular value decomposition theorem

In objective image quality metrics, one of the most important factors is the correlation of their results with the perceived quality measurements. In this paper, a new method is presented based on comparing between the structural properties of the two compared images. Based on the mathematical concept of the singular value decomposition (SVD) theorem, each matrix can be factorized to the products of three matrices, one of them related to the luminance value while the two others show the structural content information of the image. A new method to quantify the quality of images is proposed based on the projected coefficients and the left singular vector matrix of the disturbed image based on the right singular vector matrix of the original image. To evaluate this performance, many tests have been done using a widespread subjective study involving 779 images of the Live Image Quality Assessment Database, Release 2005. The objective results show a high rate of correlation with subjective quality measurements.     

位错点阵投影的三维数字成像

提出一种基于位错点阵投影的三维数字成像方法。该方法以二维点阵结构光顺序照明参考平面和被测物体表面,通过分析分别投影在参考平面和物体上的点阵的成像过程,建立投射在参考平面和被测物体表面上的点阵中同一点在成像面上横向位置错动与被测物体深度之间的数学模型。依据此模型,通过计算相对应点在成像平面上的横向位移。可计算出物体的深度图像。二维点阵可以在垂直于光轴的平面内做横向和纵向的两维移动,以填补由于离散而丢失的其他空间点的深度信息,获得高空间分辨率。实验验证了该方法的可行性,此方法对于拓扑复杂的表面和大梯度表面有较强的适应性。     

噪声对数字图像相关法计算结果的影响研究

不同的散斑场会显示出不同的灰度分布特征,并对数字图像相关方法的计算结果有着重要影响。使用计算机模拟生成一系列单向拉伸图像和双向拉伸图像,并在生成的图像中添加噪声,从而获得一系列含噪声单向拉伸图像和含噪声双向拉伸图像。使用数字图像相关方法对无噪声图像和含噪声图像分别进行相关计算,并分析研究图像应变量与相关计算结果正确率间的关系,研究结果表明发现当图像的应变量在一定范围内时,数字图像相关方法计算结果的正确率较高。同时发现,当图像的应变量较小时,噪声对相关计算结果的影响较大,随着图像应变量的增大,噪声对相关计算结果的影响逐渐减小,当图像的应变量到达一定程度时,噪声对相关计算的影响就不明显了。     

基于法向映射法的二维光纤光谱弯曲校正

二维光纤光谱图是天文望远镜系统中的光谱仪的观测结果,由此又经过一系列的后续处理,才能生成普遍意义上的一维光谱。由于光谱仪和CCD相机的光学畸变,使得二维光纤光谱图像普遍出现了弯曲现象,尤其是在光纤两端表现得尤为明显,这个问题一直以来没有很好的解决办法,也没有在任何参考文献中看到相关工作。这种弯曲给后续的抽谱工作造成困扰,处理不好会很大程度上影响波长定标,从而影响一维谱的准确性。给出的一种法向映射法可以对这种二维光纤光谱图的弯曲进行有效的校正。该法把二光纤光谱图中的每一条光纤光谱进行单独处理。首先做预处理抽取每根光纤中心线,然后把该中心线当作光滑曲线求取每个点的法线方向,以整条中心线的某个点（一般是最凸点）为基点作理想竖直线,光纤中心线所有点沿自身法线方向投影到这条竖直线上,从而实现了光纤中心线本身的校直。因光纤宽度在二维光纤光谱图中一般是7个像素,因此将光纤中心线向左向右各依次平移3个像素,分别实现上述流程,即可得到整条光纤光谱的校直结果。该法实施过程中有两个问题需要注意：一个是坐标点的均匀化问题,一个是像素灰度值的精度保持问题。坐标点的均匀化问题是由于在该法中使用了法向映射,从而造成形成后的校直线的点的疏密程度不均匀,这对后续处理不利。解决办法是在实际操作中采用三次样条插值的方法进行直线上点的密度均匀化,保证获得一系列的整数点坐标,以利于后续的处理。而像素灰度值的精度保持问题,在插值计算过程中始终保持像素灰度值的64位高精度数,最终也得到同样精度的结果,避免造成像素值损失。该法实施的最后,还需要进行头尾一致性的截取,把延伸出图像高度范围的像素去掉,仅保留图像范围内的像素点。如果没有这个过程,由于光纤本身是有宽度的,映射出来的直线长短不一,从而给后续处理造成困难。实验完整处理了整幅的二维光纤光谱图,用曲线拟合的方法较好地解决了光纤中心线提取过程中的个别地方的亮度偏差问题,用法向映射法得到了校直后的二维光谱,并做了后续的抽谱对比。在一维谱的抽谱对比中可以看到,校正前后的谱线差值在有弯曲现象存在的光谱两端变化较为明显。这个实验结果证明了该方法对光谱的弯曲情况可以完全改善--在两端的变化较大,在中间的变化很小,这完全符合观测的基本认知。不仅如此,由于像素点的错位和像素值的插值计算,也会造成流量值叠加的效果发生显著变化。因此,通过观察抽谱后的谱线在校正前后的移动情况,验证了对于特征谱线的波长位置的精确获取具有重要影响。本文创造性地设计的二维光纤光谱的自动弯曲校正的方法。     

Assessment of speckle-pattern quality in digital image correlation based on gray intensity and speckle morphology

In digital image correlation (DIC), speckle patterns are generated on the surface of a specimen to resolve uniqueness issues. Thus, speckle patterns significantly affect the accuracy of image correlation. To assess the quality of speckle patterns, the standard deviation of gray intensities within each speckle (SDGIS) is introduced as a new metric. On the basis of the cumulative distribution of SDGIS, speckle-pattern quality measurement (ρ) is proposed, which integrates the features of gray intensity and speckle morphology. Twelve speckle patterns are generated by changing the spraying time and nozzle sizes of an airbrush because these are associated with the speckle volume fraction and speckle size, respectively. In addition, three displacement fields are used to investigate the effects of speckle patterns on the accuracy of the DIC results. For the 12 speckle images associated with the three displacement fields, the correlation results demonstrate that the proposed speckle-pattern quality measurement is inversely proportional to the averaged error of the subset method. This is statistically confirmed by evaluating the correlation coefficient and p-value. Furthermore, the error of the subset method is more affected by speckle patterns than the subset size when the subset size is sufficiently large.     

物体内部三维位移场分析的数字图像相关方法

提出了物体内部三维位移场的数字图像相关分析方法，对物体变形前后，或连续变形的两个相邻状态的内部三维结构的数字图像，通过相关运算获得三维位移场.文中给出了三维相关法的体搜索窗口、相关函数及亚像素运算的相关系数拟合函数.数字模拟结果证明了三维相关法的正确性及可靠性.位移计算精度为0.02像素. 关键词： 数字图像相关 三维相关 亚像素     

High-pass filtering and frequency spectrum correlation for larger size image recognition

A new method for larger size images recognition is proposed based on high-pass filtering and frequency spectrum correlation. The input images are first filtered by a high-pass filter. Then the filtered input images, which are overlapped rather than displaced from each other in the input plane, serve as the input images for frequency spectrum correlator. As a result, only one sharp correlation peak yields in the center of output plane when the target image matches with the reference image. More importantly, the method is especially suitable for larger size image recognition because of the input and correlation ways, which ensures better utilization of the space-bandwidth product and efficient utilization of the spatial light modulator and detector. Computer simulation results are presented to demonstrate the feasibility and effectiveness of this method.     

运动背景下的帧间稳像技术

针对运动背景下帧间稳像技术,提出了一种带运动矢量修正的灰度投影运动估计算法。采用该算法分别对当前帧和参考帧的行、列计算灰度投影序列;将当前帧投影序列局部分块,分别将每一分块与参考帧行、列投影曲线进行互相关计算,得到基于局部投影的行、列运动矢量集合;以分块区域的相关置信度为权系数衡量参数,计算每一分块的像素位移权值,从而计算某一方向帧间的加权运动矢量。实验结果表明：该方法可以使运动目标造成的影响只作用于其中若干个局部分块,而其他分块不受此影响,尽可能保证稳像的准确性。采用该方法稳像后的图像与参考帧图像的均方根误差（RMSE）值明显下降,与参考帧图像更加吻合。     

基于梯度一致性约束的多光谱/全色影像最大后验融合方法

多光谱/全色影像融合可以得到高空间分辨率的多光谱影像,在影像解译和分类等方面具有十分重要的意义。提出一种基于梯度一致性约束的遥感影像融合方法。该方法在最大后验概率框架下,通过梯度一致性约束建立理想高空间分辨率多光谱影像和全色影像之间的关系,并结合多光谱影像观测模型和Huber-Markov影像先验,构建融合目标函数,最后采用梯度下降法求解得到融合影像。本文方法在目标函数中引入了梯度一致性约束,克服了现有的同类方法受限于波段数量的缺陷,同时在求解中自适应确定每个波段的迭代步长,充分顾及了各波段的光谱特性,从而既确保了融合影像的光谱信息保真度,也提高了融合影像的空间信息融入度。通过IKONOS和WorldView-2影像对该方法进行了验证,并和GS,AIHS和AMBF等融合方法从定性和定量两方面进行了比较分析。实验结果表明,相比于其他方法,该方法可以在更好保持光谱信息的同时增强影像的空间分辨率,具有更广泛的适用范围和更佳的融合效果。