Instance Matching Research Articles

A Dual Decomposition Approach to Feature Correspondence

In this paper, we present a new approach for establishing correspondences between sparse image features related by an unknown nonrigid mapping and corrupted by clutter and occlusion, such as points extracted from images of different instances of the same object category. We formulate this matching task as an energy minimization problem by defining an elaborate objective function of the appearance and the spatial arrangement of the features. Optimization of this energy is an instance of graph matching, which is in general an NP-hard problem. We describe a novel graph matching optimization technique, which we refer to as dual decomposition (DD), and demonstrate on a variety of examples that this method outperforms existing graph matching algorithms. In the majority of our examples, DD is able to find the global minimum within a minute. The ability to globally optimize the objective allows us to accurately learn the parameters of our matching model from training examples. We show on several matching tasks that our learned model yields results superior to those of state-of-the-art methods.

Efficient template matching for multi-channel images

Template matching is a computationally intensive problem aimed at locating a template within a image. When dealing with images having more than one channel, the computational burden becomes even more dramatic. For this reason, in this paper we investigate on a methodology to speed-up template matching on multi-channel images without deteriorating the outcome of the search. In particular, we propose a fast, exhaustive technique based on the Zero-mean Normalized Cross-Correlation (ZNCC) inspired from previous work related to grayscale images. Experimental testing performed over thousands of template matching instances demonstrates the efficiency of our proposal.

Ontology Instance Matching based MPEG-7 Resource Integration

Heterogeneous multimedia contents are annotated by a sharable formal conceptualization, often called ontology, and these contents, regardless of their media, become sharable resources/instances. Integration of the sharable resources and acquisition of diverse knowledge is getting researchers’ attention at a rapid pace. In this regard, MPEG-7 standard convertible to semantic Resource Description Framework (RDF) evolves for containing structured data and knowledge sources. In this paper, the authors propose an efficient approach to integrate the multimedia resources annotated by the standard of MPEG-7 schema using ontology instance matching techniques. MPEG-7 resources are usually specified explicitly by their surrounding MPEG-7 schema entities, e.g., concepts and properties, in conjunction with other linked resources. Therefore, resource integration needed schema matching as well. In this approach, the authors obtained the schema matching using their scalable ontology alignment algorithm and collected the semantically linked resources, referred to as the Semantic Link Cloud (SLC) collectively for each of the resources. Techniques were addressed to solve several data heterogeneity: value transformation, structural transformation and logical transformation. These experiments show the strength and efficiency of the proposed matching approach.

Understanding the Generalized Median Stable Matchings

Let I be a stable matching instance with N stable matchings. For each man m, order his (not necessarily distinct) N partners from his most preferred to his least preferred. Denote the ith woman in his sorted list as p i (m). Let α i consist of the man-woman pairs where each man m is matched to p i (m). Teo and Sethuraman proved this surprising result: for i=1 to N, not only is α i a matching, it is also stable. The α i ’s are called the generalized median stable matchings of I. Determining if these stable matchings can be computed efficiently is an open problem. In this paper, we present a new characterization of the generalized median stable matchings that provides interesting insights. It implies that the generalized median stable matchings in the middle—α (N+1)/2 when N is odd, α N/2 and α N/2+1 when N is even—are fair not only in a local sense but also in a global sense because they are also medians of the lattice of stable matchings. We then show that there are some families of SM instances for which computing an α i is easy but that the task is NP-hard in general. Finally, we also consider what it means to approximate a median stable matching and present results for this problem.

Consensus fingerprint matching with genetically optimised approach

Fingerprint matching has been approached using various criteria based on different extracted features. However, robust and accurate fingerprint matching is still a challenging problem. In this paper, we propose an improved integrated method which operates by first suggesting a consensus matching function, which combines different matching criteria based on heterogeneous features. We then devise a genetically guided approach to optimise the consensus matching function for simultaneous fingerprint alignment and verification. Since different features usually offer complementary information about the matching task, the consensus function is expected to improve the reliability of fingerprint matching. A related motivation for proposing such a function is to build a robust criterion that can perform well over a variety of different fingerprint matching instances. Additionally, by employing the global search functionality of a genetic algorithm along with a local matching operation for population initialisation, we aim to identify the optimal or near optimal global alignment between two fingerprints. The proposed algorithm is evaluated by means of a series of experiments conducted on public domain collections of fingerprint images and compared with previous work. Experimental results show that the consensus function can lead to a substantial improvement in performance while the local matching operation helps to identify promising initial alignment configurations, thereby speeding up the verification process. The resulting algorithm is more accurate than several other proposed methods which have been implemented for comparison.

A factor [formula omitted] approximation algorithm for two-stage stochastic matching problems

We introduce the two-stage stochastic maximum-weight matching problem and demonstrate that this problem is NP -complete. We give a factor 1 2 approximation algorithm and prove its correctness. We also provide a tight example to show the bound given by the algorithm is exactly 1 2 . Computational results on some two-stage stochastic bipartite matching instances indicate that the performance of the approximation algorithm appears to be substantially better than its worst-case performance.

The Financial Impact of ISO 9000 Certification in the United States: An Empirical Analysis

The ISO 9000 series of quality management systems standards, introduced in 1986, has been adopted at over 560,000 locations worldwide. Anecdotal evidence suggests that firms can achieve internal benefits such as quality or productivity improvements or that certification can help firms maintain or increase their market share, or both. Others argue that the standard is too generic to cause performance improvement but can be seen as a signal of good management. In this paper, we track financial performance from 1987 to 1997 of all publicly traded ISO 9000 certified manufacturing firms in the United States with SIC codes 2000–3999, and test whether ISO 9000 certification leads to productivity improvements, market benefits, and improved financial performance. We employ event-study methods, matching each certified firm to a control group of one or more noncertified firms in the same industry with similar precertification size and/or return on assets. We find that firms’ decision to seek their first ISO 9000 certification was indeed followed by significant abnormal improvements in financial performance, though the exact timing and magnitude of this effect depend on the specification of the control group. Three years after certification, the certified firms do display strongly significant abnormal performance under all control-group specifications. The degree to which the precise results vary across control-group specifications indicates that event studies should always include extensive sensitivity analysis, for instance matching by size and performance separately and jointly, using both single firms and portfolios as controls.

Computational Experience with an Approximation Algorithm on Large-Scale Euclidean Matching Instances

We consider a 2-approximation algorithm for Euclidean minimum-cost perfect matching instances proposed by the authors in a previous paper. We present computational results for both random and real-world instances having between 1,000 and 131,072 vertices. The results indicate that our algorithm generates a matching within 2% of optimal in most cases. In over 1,400 experiments, the algorithm was never more than 4% from optimal. For the purposes of the study, we give a new implementation of the algorithm that uses linear space instead of quadratic space, and appears to run faster in practice.

Representation and recognition of handprinted Chinese characters by string-matching

This paper describes an optical recognition system for the handprinted Chinese characters. It is based on a novel string representation method and an inductive learning scheme that allows flexible (or elastic) representation and matching of unknown character instances. The system scans a character instance from four different views to obtain its peripheral segment information. A string representation is designed for representing the peripheral information at each of the four views. This representation can be generalized to represent the variations in different instances of a character by using an inductive learning algorithm. A clustering algorithm is developed to group the learned representation of characters into clusters in a hierarchical tree structure. Finally, a two-stage recognition process based on the developed representation is described. Experimental results demonstrate that high recognition rates can be obtained using the developed method.

Stable matching with special preference patterns

The stable matching problem is that of matching two sets of agents in such a manner that no two unmatched agents prefer each other to their actual partners under the matching. In this paper we present some special kinds of preference patterns and discuss the elementary properties for stable matching instances with these specific preference patterns.

On the invariance of male optimal stable matching

The stable matching problem is that of matching two sets of agents in such a manner that no two unmatched agents prefer each other to their actual partners under the matching. In this paper, we present a set of sufficient conditions on the preference lists of any given stable matching instance, under which the optimality of the original male optimal stable matching is still preserved.

A VLSI system for string matching

The string matching problem consists of finding all the occurrences of a pattern p of length m in a text t of length n, m ≤ n, where p and t are strings over a finite alphabet Σ. We present a VLSI system for solving the string matching problem in O(log n + log | Σ|) time with O( m n log m log n) area. The system is structured as an m × n mesh of trees, in which n − m + 1 additional trees are introduced on the main diagonals of the mesh with no asymptotic increase of the area. Pipelining allows us to consecutively solve several different string matching instances with an O(log|Σ|) time period between two successive instances. Partitioning approaches are also dealt with.