Improved Dynamic Dictionary Matching

In the dynamic dictionary matching problem, a dictionary D contains a set of patterns that can change over time by insertion and deletion of individual patterns. The user also presents text strings and asks for all occurrences of any patterns in the text. The two main contributions of this paper are: (1) a faster algorithm for dynamic string dictionary matching with bounded alphabets, and (2) a dynamic dictionary matching algorithm for two-dimensional texts and patterns. The first contribution is based on an algorithm that solves the general problem of maintaining a sequence of well-balanced parentheses under the operations insert, delete, and find nearest enclosing parenthesis pair. The main new idea behind the second contribution is a novel method to efficiently manipulate failure links for two-dimensional patterns.     

Dynamic Dictionary Matching in External Memory

In thedynamic dictionary matchingproblem, a dictionaryDcontains a set of patterns that can change over time under insertion and deletion of individual patterns. Given an arbitrary textT, we must efficiently list all the dictionary patterns that occur at each text position. We investigate the I/O complexity of this problem for a large dictionary that must be stored in external storage devices. By following a completely new approach, we devise an efficient solution which is based upon the SB-tree data structure (P. Ferragina and R. Grossi, 1995,in“Proc. ACM Symposium on Theory of Computing,” pp. 693–702), and a novel notion of certificate for the dictionary matching problem. Our data structure can be adapted to efficiently work in main memory and to solve other problems, thus providing a new insight into the nature of the dictionary matching problem.     

二维逐步正交匹配追踪算法

在图像等二维信号的应用与处理上,常规压缩感知理论框架存在重构算法效果差、图像块效应明显、对噪声敏感等问题。针对这些问题,根据现有二维观测模型和二维重构算法设计思想,可以设计一种新的重构算法：二维逐步正交匹配追踪算法。该算法借鉴了相关一维重构算法的设计思想,通过每次迭代选取符合阈值条件的多列原子进而正交化处理的步骤,提升了重构效率,改善了恢复图像质量。理论分析和实验结果表明,提出的算法在重构时间得到控制的情况下,得到的图像信噪比有较大提升,超越了现有典型的二维重构算法。     

基于形状轮廓特征的金字塔匹配算法

形状检索在计算机视觉中一直是一个具有挑战性的问题，其中对形状特征直方图距离的测量是评价形状检索算法优劣的一个重要因素。针对轮廓特征的直方图距离测量，算法引进一种在图像分类领域中应用广泛的金字塔匹配算法。不同于其他传统的直方图度量算法，金字塔匹配算法将形状的轮廓分成若干块，给每一块分配相应的权重，然后分别统计块中的特征，再计算特征的加权和进行相似度的测量。通过在不同形状数据集下实验，该方法能够有效地进行形状匹配和检索，且能得到较好的形状匹配精度。     

Efficient Algorithms for Image Template and Dictionary Matching

Given a large text image and a small template image, the Template Matching Problem is that of finding every location within the text which looks like the pattern. This problem, which has received attention for low-level image processing, has been formalized by defining a distance metric between arrays of pixels and finding all subarrays of the large image which are within some threshold distance of the template. These so-called metric methods tends to be too slow for many applications, since evaluating the distance function can take too much time. We present a method for quickly eliminating most positions of the text from consideration as possible matches. The remaining candidate positions are then evaluated one by one against the template for a match. We are still guaranteed to find all matching positions, and our method gives significant speed-ups. Finally, we consider the problem of matching a dictionary of templates against a text. We present methods which are much faster than matching the templates individually against the input image.     

基于Hash结构词典的双向最大匹配分词法

针对当前自然语言处理中中文分词基于词典的机械分词方法,正序词典不能作为逆向最大匹配分词词典以及反序词典维护困难的问题,提出一种新的词典构造方法并设计了相应的双向最大匹配算法,同时在算法中加入了互信息歧义处理模块来处理分词中出现的交集型歧义。该算法可以在分词的过程中显著提高分词的精确度,适用于对词语切分精度要求较高的中文语言处理系统。     

基于2D轮廓图的三维模型相似性比较研究

将三维模型分别从正视、侧视、俯视3个角度投影到3个平面.再将3个平面上的2D投影轮廓图等分成若干个扇形区域.计算每个区域内的顶点与模型重心的最大距离.如此分别得到三个平面上2D轮廓图的特征向量.提取每个模型的这3个特征向量作为其形状描述符,然后进行接下来的模型相似性匹配.实验结果表明:该方法不但简便而且具有较好的三维模型检索准确性.     

基于极坐标区间运算的2D形状匹配

形状匹配是遥感图像目标识别、字符识别、手形识别和步态识别等任务中的关键步骤之一.针对刚体识别任务中形状匹配易受方向、尺度和位置等仿射变化量影响的情况,提出了一种新的基于极坐标区间运算的2D形状匹配算法.该算法首先以形状区域的中心点为极点,区域的最长轴方向为极轴,对形状区域进行归一化的极坐标变换;然后定义了同一角度对应的区域内点区间之间的运算;最后定义了两个区域归一化极坐标变换结果在区间运算下的相似度函数,用以表征两个区域之间的匹配度.从可见光遥感图像中提取的实物图像实验结果证明,该方法能够有效归类相似形状,并能区分各类不同的形状.     

Succinct nearest neighbor search

An efficient and universal similarity search solution is a holy grail for multimedia information retrieval. Most similarity indexes work by mapping the original multimedia objects into simpler representations, which are then searched by proximity using a suitable distance function.     

基于亚像素块匹配和字典学习的超分辨率算法

针对基于字典学习算法的计算效率低,且大多局限于处理单帧图像的问题,提出了一种基于亚像素块匹配和字典学习的超分辨算法,以实现对多帧图像的重构。采用亚像素块匹配方法对图像进行配准,依据配准结果构造低分辨率字典,并通过计算辅助图像块与目标图像块的相似度来选择用于重构的图像块。在Matlab平台上,将该算法用于静态图像和视频图像处理,获得了较好的重构效果。     

Succinct Representation of Labeled Graphs

In many applications, the properties of an object being modeled are stored as labels on vertices or edges of a graph. In this paper, we consider succinct representation of labeled graphs. Our main results are the succinct representations of labeled and multi-labeled graphs (we consider planar triangulations, planar graphs and k-page graphs) to support various label queries efficiently. The additional space cost to store the labels is essentially the information-theoretic minimum. As far as we know, our representations are the first succinct representations of labeled graphs. We also have two preliminary results to achieve the main contribution. First, we design a succinct representation of unlabeled planar triangulations to support the rank/select of edges in ccw (counter clockwise) order in addition to the other operations supported in previous work. Second, we design a succinct representation for a k-page graph when k is large to support various navigational operations more efficiently. In particular, we can test the adjacency of two vertices in O(lg?k) time, while previous work uses O(k) time.     

Succinct representation of dynamic trees

We study the problem of maintaining a dynamic ordered tree succinctly under updates of the following form: insertion or deletion of a leaf, insertion of a node on an edge (edge subdivision) or deletion of a node with only one child (the child becomes a child of its former grandparent). We allow satellite data of a fixed size to be associated to the nodes of the tree.We support update operations in constant amortized time and support access to satellite data and basic navigation operations in worst-case constant time; the basic navigation operations include parent, first/last-child, previous/next-child. These operations are moving from a node to its parent, leftmost/rightmost child, and its previous and next child respectively.We demonstrate that to efficiently support more extended operations, such as determining the i-th child of a node, rank of a child among its siblings, or size of the subtree rooted at a node, one requires a restrictive pattern for update strategy, for which we propose the finger-update model. In this model, updates are performed at the location of a finger that is only allowed to crawl on the tree between a child and a parent or between consecutive siblings. Under this model, we describe how the named extended operations are performed in worst-case constant time.Previous work on dynamic succinct trees (Munro et al., 2001 [17]; Raman and Rao, 2003 [19]) is mainly restricted to binary trees and achieves poly-logarithmic (Munro et al., 2001 [17]) or “poly-log-log” (Raman and Rao, 2003 [19]) update time under a more restricted model, where updates are performed in traversals starting at the root and ending at the root and queries can be answered when the traversal is completed. A previous result on ordinal trees achieves only sublinear amortized update time and “poly-log-log” query time (Gupta et al., 2007 [11]). More recently, the update time has been improved to O(logn/loglogn) while queries can be performed in O(logn/loglogn) time (Sadakane and Navarro, 2010 [20]).     

结合全色度HEVC和有损字典算法的屏幕图像编码

鉴于传统视频压缩算法对屏幕视频编码的效果不理想,提出了一种误差限自适应的有损字典编码方案,并和全色度HEVC有机融合,形成LDSC(Lossy Dual-coder Single Chroma-sampling-rate)算法。该算法根据屏幕内容中不同类型区域的编码特性,自适应地选择率失真性能较好的编码结果放入码流。为了评价有损字典编码的性能,提出了基于匹配长度和匹配失真的LD-Cost(Length Distortion Cost)模型,并使用拉格朗日乘子法对模型进行分析和计算。对于有损字典编码过程中可能出现的误差积累现象,给出了可行的解决方案。实验结果表明,与基于无损字典编码的双编码器方案相比,在不损伤主观图像质量的前提下,LDSC算法能明显提高编码性能,对连续色调内容居多的屏幕视频也表现出更好的编码适应性,全I帧配置下BD-rate性能比基于无损字典编码的方案提高了3%~15%,与单纯使用HEVC相比,BD-rate性能提高了7%~49%。     

一种新的二维开曲线匹配算法研究

在提取碎片轮廓的基础上,提出了一种基于相似变换下的新的尺寸不变为标示符的二维开曲线匹配方法。基本思想是首先以弧长的曲率绝对值的积分方法,通过对轮廓重采样来计算轮廓曲线上的特征点,特征点分曲线为若干段,然后特征段之间的Hausdorff距离来比较两曲线的段的相似性,当Hausdorff距离小于给定的容差时,可认为相应的轮廓是匹配的,实验证明算法更快有效。     

平面点匹配的一点校准算法

点模式匹配是一项重要的视觉课题。对于一个平面点集,由平移和旋转并伴有一定噪声作用产生另一点集,提出一个基于一点校准的点模式快速匹配算法,并推广到带有属性点的匹配问题中。基于一点校准的点模式匹配算法,其计算复杂性为O(mn),其中m,n分别是两个点集所含点的个数,比基于两点距离近似相等的校准匹配算法,其计算复杂性为O(m2nl)(其中l为第二个点集中与第一个点集中任两个点的距离近似相等的平均个数),极大地减少了计算量。