Graph-based representation for similarity retrieval of symbolic images

Image retrieval from an image database by the image objects and their spatial relationships has emerged as an important research subject in these decades. To retrieve images similar to a given query image, retrieval methods must assess the similarity degree between a database image and the query image by the extracted features with acceptable efficiency and effectiveness. This paper proposes a graph-based model SRG (spatial relation graph) to represent the semantic information of the contained objects and their spatial relationships in an image with no file annotation. In an SRG graph, the image objects are symbolized by the predefined class names as vertices and the spatial relations between object pairs are represented as arcs. The proposed model assesses the similarity degree between two images by calculating the maximum common subgraph of two corresponding SRG’s through intersection, which has quadratic time complexity owing to the characteristics of SRG. Its efficiency remains quadratic regardless of the duplication rate of the object symbols. The extended model SRG_T is also proposed, with the same time complexity, for the applications that need to consider the topological relations among objects. A synthetic symbolic image database and an existing image dataset are used in the conducted experiments to verify the performance of the proposed models. The experimental results show that the proposed models have compatible retrieval quality with remarkable efficiency improvements compared with three well-known methods LCS_Clique, SIM_R, and 2D Be-string, where LCS_Clique utilizes the number of objects in the maximum common subimage as its similarity function, SIM_R uses accumulation-based similarity function of similar object pairs, and 2D Be-string calculates the similarity of 2D patterns by the linear combination of two 1D similarities.     

Speeding up transposition-invariant string matching

Finding the longest common subsequence (LCS) of two given sequences A=a₀a₁…am−1 and B=b₀b₁…bn−1 is an important and well studied problem. We consider its generalization, transposition-invariant LCS (LCTS), which has recently arisen in the field of music information retrieval. In LCTS, we look for the LCS between the sequences A+t=(a₀+t)(a₁+t)…(am−1+t) and B where t is any integer. We introduce a family of algorithms (motivated by the Hunt-Szymanski scheme for LCS), improving the currently best known complexity from O(mnloglogσ) to O(Dloglogσ+mn), where σ is the alphabet size and D?mn is the total number of dominant matches for all transpositions. Then, we demonstrate experimentally that some of our algorithms outperform the best ones from literature.     

Spatial inference and similarity retrieval of an intelligent image database system based on object's spanning representation

In this paper, we presented a novel image representation method to capture the information about spatial relationships between objects in a picture. Our method is more powerful than all other previous methods in terms of accuracy, flexibility, and capability of discriminating pictures. In addition, our method also provides different degrees of granularity for reasoning about directional relations in both 8- and 16-direction reference frames. In similarity retrieval, our system provides twelve types of similarity measures to support flexible matching between the query picture and the database pictures. By exercising a database containing 3600 pictures, we successfully demonstrated the effectiveness of our image retrieval system. Experiment result showed that 97.8% precision rate can be achieved while maintaining 62.5% recall rate; and 97.9% recall rate can be achieved while maintaining 51.7% precision rate. On an average, 86.1% precision rate and 81.2% recall rate can be achieved simultaneously if the threshold is set to 0.5 or 0.6. This performance is considered to be very good as an information retrieval system.     

A new relevance feedback technique for iconic image retrieval based on spatial relationships

Due to the popularity of Internet and the growing demand of image access, the volume of image databases is exploding. Hence, we need a more efficient and effective image searching technology. Relevance feedback technique has been popularly used with content-based image retrieval (CBIR) to improve the precision performance, however, it has never been used with the retrieval systems based on spatial relationships. Hence, we propose a new relevance feedback framework to deal with spatial relationships represented by a specific data structure, called the 2D B_e-string. The notions of relevance estimation and query reformulation are embodied in our method to exploit the relevance knowledge. The irrelevance information is collected in an irrelevant set to rule out undesired pictures and to expedite the convergence speed of relevance feedback. Our system not only handles picture-based relevance feedback, but also deals with region-based feedback mechanism, such that the efficacy and effectiveness of our retrieval system are both satisfactory.     

A fast algorithm for computing a longest common increasing subsequence

Let A=〈a₁,a₂,…,am〉 and B=〈b₁,b₂,…,bn〉 be two sequences, where each pair of elements in the sequences is comparable. A common increasing subsequence of A and B is a subsequence 〈ai₁=bj₁,ai₂=bj₂,…,ail=bjl〉, where i₁<i₂<?<il and j₁<j₂<?<jl, such that for all 1?k<l, we have aik<aik+1. A longest common increasing subsequence of A and B is a common increasing subsequence of the maximum length. This paper presents an algorithm for delivering a longest common increasing subsequence in O(mn) time and O(mn) space.     

一种基于形状的图像信息检索方法

该文把一幅图像看成是由一些区域构成的,这些区域在其内部有着颜色或纹理等方面的相似性,图像的形状由这些区域的边界线及其空间关系来描述.要根据图像中所包含的物体或场景在形状方面的特征进行检索,关键问题是形状相似性的度量及其空间关系的表示与匹配.文章应用变形模板匹配技术,提出了较为合理的简单形状相似性计算方法,而这些简单形状之间的空间关系则由二维集合串来表示.文章还给出了空间关系匹配算法,在检索方法上,将整个检索过程分为初级检索、检索求精与空间关系匹配3个阶段.实验表明,此方法既有较高的检索速度,又有较高的检索精度.     

Applying VSM and LCS to develop an integrated text retrieval mechanism

Text retrieval has received a lot of attention in computer science. In the text retrieval field, the most widely-adopted similarity technique is using vector space models (VSM) to evaluate the weight of terms and using Cosine, Jaccard or Dice to measure the similarity between the query and the texts. However, these similarity techniques do not consider the effect of the sequence of the information. In this paper, we propose an integrated text retrieval (ITR) mechanism that takes the advantage of both VSM and longest common subsequence (LCS) algorithm. The key idea of the ITR mechanism is to use LCS to re-evaluate the weight of terms, so that the sequence and weight relationships between the query and the texts can be considered simultaneously. The results of mathematical analysis show that the ITR mechanism can increase the similarity on Jaccard and Dice similarity measurements when a sequential relationship exists between the query and the texts.     

Spatial reasoning and similarity retrieval for image database systems based on RS-strings

In this paper, we propose a rotation-invariant spatial knowledge representation called RS-string. Then we present the string generation algorithm to automatically generate RS-strings for segmented pictures. We also propose the spatial reasoning and similarity retrieval algorithms based on RS-strings. The similarity retrieval algorithm is much more flexible than all previous 2D string representations because our approach can consider every possible view of a query picture. Thus the system does not require the user to provide a query picture which must have the same orientation as that of a database picture. Finally, we provide several examples to demonstrate the capabilities of spatial reasoning and similarity retrieval based on the RS-string representation.     

A New Efficient Algorithm for Computing the Longest Common Subsequence

The Longest Common Subsequence (LCS) problem is a classic and well-studied problem in computer science. The LCS problem is a common task in DNA sequence analysis with many applications to genetics and molecular biology. In this paper, we present a new and efficient algorithm for solving the LCS problem for two strings. Our algorithm runs in O(ℛlog log n+n) time, where ℛ is the total number of ordered pairs of positions at which the two strings match. Preliminary version appeared in [24]. C.S. Iliopoulos is supported by EPSRC and Royal Society grants. M.S. Rahman is supported by the Commonwealth Scholarship Commission in the UK under the Commonwealth Scholarship and Fellowship Plan (CSFP) and is on Leave from Department of CSE, BUET, Dhaka-1000, Bangladesh.     

Beam search for the longest common subsequence problem

The longest common subsequence problem is a classical string problem that concerns finding the common part of a set of strings. It has several important applications, for example, pattern recognition or computational biology. Most research efforts up to now have focused on solving this problem optimally. In comparison, only few works exist dealing with heuristic approaches. In this work we present a deterministic beam search algorithm. The results show that our algorithm outperforms the current state-of-the-art approaches not only in solution quality but often also in computation time.     

Variants of constrained longest common subsequence

We consider a variant of the classical Longest Common Subsequence problem called Doubly-Constrained Longest Common Subsequence (DC-LCS). Given two strings s₁ and s₂ over an alphabet Σ, a set Cs of strings, and a function Co:Σ→N, the DC-LCS problem consists of finding the longest subsequence s of s₁ and s₂ such that s is a supersequence of all the strings in Cs and such that the number of occurrences in s of each symbol σ∈Σ is upper bounded by Co(σ). The DC-LCS problem provides a clear mathematical formulation of a sequence comparison problem in Computational Biology and generalizes two other constrained variants of the LCS problem that have been introduced previously in the literature: the Constrained LCS and the Repetition-Free LCS. We present two results for the DC-LCS problem. First, we illustrate a fixed-parameter algorithm where the parameter is the length of the solution which is also applicable to the more specialized problems. Second, we prove a parameterized hardness result for the Constrained LCS problem when the parameter is the number of the constraint strings (|Cs|) and the size of the alphabet Σ. This hardness result also implies the parameterized hardness of the DC-LCS problem (with the same parameters) and its NP-hardness when the size of the alphabet is constant.     

New efficient algorithms for the LCS and constrained LCS problems

In this paper, we study the classic and well-studied longest common subsequence (LCS) problem and a recent variant of it, namely the constrained LCS (CLCS) problem. In the CLCS problem, the computed LCS must also be a supersequence of a third given string. In this paper, we first present an efficient algorithm for the traditional LCS problem that runs in O(Rloglogn+n) time, where R is the total number of ordered pairs of positions at which the two strings match and n is the length of the two given strings. Then, using this algorithm, we devise an algorithm for the CLCS problem having time complexity O(pRloglogn+n) in the worst case, where p is the length of the third string.     

An almost-linear time and linear space algorithm for the longest common subsequence problem

There are two general approaches to the longest common subsequence problem. The dynamic programming approach takes quadratic time but linear space, while the nondynamic-programming approach takes less time but more space. We propose a new implementation of the latter approach which seems to get the best for both time and space for the DNA application.     

Most discriminating segment – Longest common subsequence (MDSLCS) algorithm for dynamic hand gesture classification

In this work, we consider the recognition of dynamic gestures based on representative sub-segments of a gesture, which are denoted as most discriminating segments (MDSs). The automatic extraction and recognition of such small representative segments, rather than extracting and recognizing the full gestures themselves, allows for a more discriminative classifier. A MDS is a sub-segment of a gesture that is most dissimilar to all other gesture sub-segments. Gestures are classified using a MDSLCS algorithm, which recognizes the MDSs using a modified longest common subsequence (LCS) measure. The extraction of MDSs from a data stream uses adaptive window parameters, which are driven by the successive results of multiple calls to the LCS classifier. In a preprocessing stage, gestures that have large motion variations are replaced by several forms of lesser variation. We learn these forms by adaptive clustering of a training set of gestures, where we reemploy the LCS to determine similarity between gesture trajectories. The MDSLCS classifier achieved a gesture recognition rate of 92.6% when tested using a set of pre-cut free hand digit (0–9) gestures, while hidden Markov models (HMMs) achieved an accuracy of 89.5%. When the MDSLCS was tested against a set of streamed digit gestures, an accuracy of 89.6% was obtained. At present the HMMs method is considered the state-of-the-art method for classifying motion trajectories. The MDSLCS algorithm had a higher accuracy rate for pre-cut gestures, and is also more suitable for streamed gestures. MDSLCS provides a significant advantage over HMMs by not requiring data re-sampling during run-time and performing well with small training sets.     

一种XML文档结构相似度计算方法

对XML文档树路径模型进行扩展,加入了路径的频率信息.基于此路径-频率模型,提出一种带有位置仅重的基于路径的结构相似度计算方法(WLCS),并在此基础上提出基于路径频率的XML文档结构向量化方法.在真实数据集上的实验结果表明,WLCS方法召回率和准确率均高于当前存在的基于路径计算相似度的方法,适合于对来自不同DTD的XML文档的相似度比较.     

Doubly-Constrained LCS and Hybrid-Constrained LCS problems revisited

We revisit two recently studied variants of the classic Longest Common Subsequence (LCS) problem, namely, the Doubly-Constrained LCS (DC-LCS) and Hybrid-Constrained LCS (HC-LCS) problems. We present finite automata based algorithms for both problems.