交互式立体显示中基于图像拼合的视点合成

在立体显示中,视点合成是实现交互性的关键技术,即在三维(3D)场景中通过自由选择视点而获得环视能力.本文将视点插值和基于图像拼合的视点变形技术相结合,提出一种中间视合成算法.首先均匀化原始立体图像对;然后只对前景对象区域进行视差估计以提高视差匹配的速度和精度;接着确定左右视点中的可靠区域,根据可靠区域生成过渡中间视点;最后,采用视点插值结合变形的方法,由过渡视合成中间视点.实验结果表明合成的中间视点图像质量良好,而且合成速度也明显提高.本文算法可用于实时 3D 视频应用的交互式立体显示,可以实现任意视点的实时绘制.     

一种基于边缘分割的全局立体匹配算法

匹配基元的选取对立体匹配的视差计算精度与立体匹配的速度有着重要影响.首先提出了一种利用边缘分割扫描线所得到的线段作为匹配基元的方法,采用基于动态阈值的Canny检测算子提取边缘,将扫描线分割为线段,使得分割线段的长度可以根据图像的状况进行调整,从而在保证视差计算精度的同时有效地提高了速度.另外,以分割后的线段作为结点构建生成树(ST),使用树型动态规划算法计算基元的视差值,通过增加扫描线之间的约束,从而得到全局最优视差.实验表明这种算法具有较高的视差计算精度和较好的实时性.     

遮挡检测/立体匹配中的分段动态规划法

 为了显式地在视差图上标记出遮挡区域,本文在计算视差空间的基础上,利用动态规划算法搜索出最佳视差曲线.计算所得的视差曲线上有三种状态标记:匹配状态和二种遮挡状态.为了保证视差曲线通过路径控制点,提出了一种分段式动态规划算法.算法将视差空间影像划分为路径控制区和非路径控制区.在路径控制区强制路径通过路径控制点,在非路径控制区采用动态规划进行路径最佳搜索.为保证路径控制点高度可靠,提出了选择路径控制点的4个准则.实验结果表明,新算法比传统的动态规划算法在遮挡检测和匹配精度上都有一定的提高,算法可靠性强,运算量小.     

基于改进代价计算和视差候选策略的立体匹配

立体匹配算法在图像弱纹理区和重复纹理区存在匹配困难、误差大的问题,为此提出一种基于改进代价计算和视差候选策略的立体匹配算法。首先结合改进的Census变换和自适应加权融合的双向梯度信息来计算初始匹配代价,提高代价计算的可靠性。其中:为传统Census变换增加内圈编码,提高邻域信息利用率,同时降低噪声的影响;利用自适应权重函数融合横向和纵向梯度代价,降低物体边缘区域的误匹配率。其次,采用自适应十字交叉窗口进行代价聚合,并通过建立候选视差集和引入邻域视差信息的方法来获取初始视差。最后通过两轮插值策略优化视差。实验结果表明,所提算法能够提高弱纹理区和重复纹理区的匹配效果,在Middlebury中4幅标准立体图像对的平均误匹配率为5.33%。     

图像多通道边缘信息辅助的快速立体匹配算法

提出一种基于图像边缘及色彩信息辅助的立体匹配算法.首先给出一种彩色边缘侦测算子来获取彩色图像边缘信息,利用基于归一化相关系数的极线立体匹配得到视差图像边缘像素的视差信息.然后利用已获取彩色图像对的边缘信息对双步骤动态规划匹配策略进行匹配指导,获取稠密视差图并进一步优化.实验表明,提出的算法能很好地降低图像对中强度接近物体的误匹配概率的同时,有效提高匹配算法的运行效率.     

基于公共点提取的多视图像校正

自由视点立体视频既可以给观看者强烈的立体感,又可以为用户同时呈现不同角度的视点画面。但是垂直视差的存在,会影响合成立体图像的质量,出现图像的模糊,严重的垂直视差会使图像出现“跳跃”现象,甚至不能观看。因此,多视点图像的校正是提高3DTV观看质量的重要步骤之一。现阶段关于双视点图像校正技术的研究已较多,但多视点图像校正研究却需深入。其中,基于多相机标定参数的校正算法需要提供预先的标定数据,应用不便。本文提出了一种在无多摄像机标定参数情况下,消除垂直视差的校正方法。首先,使用SIFT算法提取图像特征点。然后,利用多视点图像公共点的匹配关系,选取唯一基准视点图像平面计算各视点校正矩阵。实验结果表明,该方法有效降低了垂直视差,限制了校正误差的传播,从而保持了校正后图像质量的一致性。      

视点相关光线投射算法在三维云可视化方面的应用

针对绘制海量云数据时传统光线投射算法计算量大、绘制缓慢等问题,提出了一种基于视点相关光线投射算法的三维云可视化方法。该方法首先采用基于多分辨率方法的数据预处理策略,然后采用基于视点相关系数的自适应采样频率计算方法确定重采样点的位置,并利用三阶段插值算法计算重采样点的值,最后采用基于块重要性加权香农熵的二维传输函数完成颜色值和不透明度的映射关系,从而实现图像的合成。实验结果表明,该方法计算复杂度低、执行效率高,与现有的光线投射算法相比,不仅图像的重构质量得到了提高,绘制时间也减少了约30%。     

基于互信息的亚像素级立体视觉点匹配方法研究

在立体视觉测量中,为获得更高精度,往往需要将视差计算精确到亚像素级。将互信息理论引入双目图像配准并结合多分辨率技术实现亚像素级点匹配。采用Bouguet立体校正算法对左右图像进行极线校正,利用Harris角点探测器检测目标并将获取角点作为待匹配点,采用最大互相关法进行搜索确定像素级匹配点。然后对以左右匹配点为中心的20×20邻域图像进行插值并分别放大10倍和100倍,采用互信息方法先对低分辨率图像进行配准,再在高分辨率图像上进一步细化求精,结合像素级匹配的整数视差可得最终亚像素级视差。实验结果表明,该方法能将视差精度提高到0.01像素。     

多分辨率与视差梯度相结合的快速区域匹配

针对传统区域匹配算法中存在大量冗余计算,提出了一种多分辨率与视差梯度相结合的快速匹配算法。首先通过均值采样得到低分辨率图像对,再通过改进的视差梯度约束算法进行匹配得到低分辨率图像的视差图,最后引入盒滤波思想指导原始视差图的生成。实验表明,该快速匹配算法相对于传统的区域匹配算法,在改善精度的同时,减少了三倍以上运算时间。     

基于改进匹配代价和均值分割的最小生成树立体匹配算法

尽管传统的立体匹配模型在精度和鲁棒性方面都表现出了良好的性能，但在弱纹理和深度不连续区域的视差精度问题依然存在。针对上述问题，提出了一种基于改进匹配代价和均值分割的最小生成树立体匹配算法。首先，在匹配代价计算阶段，通过Census变换进行初始匹配代价计算，利用Sobel算子对输入图像进行边缘信息提取，将提取后的图像边缘信息与Census变换后的匹配代价值进行融合，并将其与基于图像亮度信息的代价值进行非线性融合，以提高匹配代价的精度；然后，使用最小生成树代价聚合模型进行聚合操作并利用赢者通吃策略估计图像的初始视差；最后，在视差优化阶段，采用MeanShift算法对图像进行分割，结合图像的轮廓信息对误匹配点进行修正，进一步提高在弱纹理及边缘区域的视差精度。实验结果表明，与一些传统算法相比，所提算法具有更高的视差精度，且视差图的边缘、纹理较其他算法更为平滑，具有更强的鲁棒性。     

MULTI-EPIPOLAR LINES MATCHING-BASED RAY-SPACE INTERPOLATION FOR FREE VIEWPOINT VIDEO SYSTEM

Ray-space based arbitrary viewpoint rendering without complex object segmentation or model construction is the main technology to realize Free Viewpoint Video （FVV） system for complex scenes. Ray-space interpolation and compression are two key techniques for the solution. In this paper, correlation among multiple epipolar lines in ray-space data is analyzed, and a new method of ray-space interpolation with multi-epipolar lines matching is proposed. Comparing with the pixel-based matching interpolation method and the block-based matching interpolation method, the proposed method can achieve higher Peak Signal to Noise Ratio （PSNR） in interpolating ray- space data and rendering arbitrary viewpoint images.     

On determining the independent point set for doubly periodic arrays and encoding two-dimensional cyclic codes and their duals

For the purpose of encoding two-dimensional cyclic (TDC) codes, an effective algorithm for finding the independent point set of an arbitrary module of doubly periodic (DP) arrays is proposed. In addition, a method for determining the characteristic ideal of a given set of DP arrays is exhibited. With the aid of these methods it is possible to specify the structure of the generator and check ideals of TDC codes. By applying this algorithm to nonsemisimple binary TDC codes with small areas, several optimal linear codes have been found and their previously unknown TDC structures have been exhibited.     

Dynamic programming based adaptive step integration method for maneuvering fluctuating target detection

An adaptive long time integration method based on dynamic programming (DP) is proposed for the detection of high speed maneuvering fluctuating targets. The proposed method is aimed at detecting a target with an unpredictable range migration and fluctuating echoes by jointly applying three main ideas: the improved DP procedure that could search the maneuver position and velocity at each frame; the multi-pulse integration that could suppress the influence of fluctuation; and the adaptive step with fading factor that could allow the integration time to be suitable for each searching velocity. Compared with the existing methods, the target energy could be integrated along its trajectory using the proposed method without estimating the specific motion parameters, which makes the proposed method applicable to a target with arbitrary motion. Simulation results and performance comparisons show the superiority of the proposed method.     

Overview of free viewpoint television

We have been developing ray-based 3D information systems that consist of ray acquisition, ray processing, and ray display. Free viewpoint television (FTV) based on the ray-space method is a typical example. FTV will bring an epoch-making change in the history of television because it enables us to view a distant 3D world freely by changing our viewpoints as if we were there. We constructed a real-time FTV including the complete chain from capturing to display. A new algorithm was developed to generate free viewpoint images. In addition, a new user interface is presented for FTV to make full use of 3D information. FTV is not a pixel-based system but a ray-based system. We are creating ray-based image engineering through the development of FTV.     

Coherent integration detection method for maneuvering target based on dynamic programming

In allusion to the coupling relationship among radial velocity variation, range migration and Doppler spread of a maneuvering target, a dynamic programming (DP) based long time coherent integration (DPCI) detection method is proposed. The energy of the target of high velocity and strong maneuver could be integrated along its trajectory by the proposed DPCI method with high efficiency. The main idea of the DPCI is the joint execution of two operations: the modified DP procedure that could search the current position and velocity of the target, and the improved coherent integration that could calibrate the Doppler shift during the entire process with any large phase spread. The proposed method is applicable to a target with arbitrary motion without estimating its specific movement parameters. Simulation results and performance comparisons show the accuracy and superiority of the proposed method.     

General theory of doubly periodic arrays over an arbitrary finite field and its applications

A general theory of doubly periodic (DP) arrays over an arbitrary finite field GF(q)is presented. First the basic properties of DP arrays are examined. Next modules of linear recurring (LR) arrays are defined and their algebraic properties discussed in connection with ideals in an extension ringtilde{R}of the ringRof bivariate polynomials with coefficients in GF(q). A finitetilde{R}-module of DP arrays is shown to coincide with thetilde{R}-module of LR arrays dermed by a zero-dimensional ideal intilde{R}. Equivalence relations between DP arrays are explored, i.e., rearrangements of arrays by means of unimodular transformations. Decimation and interleaving of arrays are defined in a two-dimensional sense. The general theory is followed by application to irreducible LR arrays. Among irreducible arrays,M-arrays are a two-dimensional analog ofM-sequences and may be constructed fromM-sequences by means of unimodular transformations. The results of this paper are also important in studying properties of Abelian codes.     

基于分区动态规划和航迹关联的弱小目标检测

序列图像中低信噪比运动弱小目标的实时检测算法，是精确制导系统中的关键算法之一。提出一种图像序列弱小目标实时检测新算法，采用分方向区间的动态规划算法和二值图像航迹关联检测。通过在动态规划能量累加过程中引入方向限制，减小了噪声轨迹能量积累和目标轨迹能量扩散，提高了算法的目标检测能力。对算法的检测性能进行了仿真实验，结果表明该算法能有效检测深空背景下信噪比大于1.8、运动方向任意、速度不大于1像素/帧的多个运动弱小目标。     

Full-rate full-diversity space-frequency codes with optimum coding advantage

A general space-frequency (SF) block code structure is proposed that can guarantee full-rate (one channel symbol per subcarrier) and full-diversity transmission in multiple-input multiple-output-orthogonal frequency-division multiplexing (MIMO-OFDM) systems. The proposed method can be used to construct SF codes for an arbitrary number of transmit antennas, any memoryless modulation and arbitrary power-delay profiles. Moreover, assuming that the power-delay profile is known at the transmitter, we devise an interleaving method to maximize the overall performance of the code. We show that the diversity product can be decomposed as the product of the "intrinsic" diversity product, which depends only on the used signal constellation and the code design, and the "extrinsic" diversity product, which depends only on the applied interleaving method and the power delay profile of the channel. Based on this decomposition, we propose an interleaving strategy to maximize the "extrinsic" diversity product. Extensive simulation results show that the proposed SF codes outperform the previously existing codes by about 3-5 dB, and that the proposed interleaving method results in about 1-3-dB performance improvement compared to random interleaving.     

Music Analysis Using Hidden Markov Mixture Models

We develop a hidden Markov mixture model based on a Dirichlet process (DP) prior, for representation of the statistics of sequential data for which a single hidden Markov model (HMM) may not be sufficient. The DP prior has an intrinsic clustering property that encourages parameter sharing, and this naturally reveals the proper number of mixture components. The evaluation of posterior distributions for all model parameters is achieved in two ways: 1) via a rigorous Markov chain Monte Carlo method; and 2) approximately and efficiently via a variational Bayes formulation. Using DP HMM mixture models in a Bayesian setting, we propose a novel scheme for music analysis, highlighting the effectiveness of the DP HMM mixture model. Music is treated as a time-series data sequence and each music piece is represented as a mixture of HMMs. We approximate the similarity of two music pieces by computing the distance between the associated HMM mixtures. Experimental results are presented for synthesized sequential data and from classical music clips. Music similarities computed using DP HMM mixture modeling are compared to those computed from Gaussian mixture modeling, for which the mixture modeling is also performed using DP. The results show that the performance of DP HMM mixture modeling exceeds that of the DP Gaussian mixture modeling.     

一种基于动态规划法的雷达微弱多目标检测方法

在雷达微弱信号检测领域中,动态规划法是一种有效的能量积累方法.本文首先分析了应用动态规划法对多个微弱目标信号积累后数据图像的特点,在此基础上通过求极值实现了对多目标的检测.然后对算法的虚警概率和检测概率做了理论分析,同时进行了仿真实验.仿真结果表明,与传统的只作相参积累的检测方法相比,基于动态规划的极值法带来了4~5dB的非相参积累增益.