Graph Matching Based on Node Signatures

Abstract

Abstract. We present an algorithm for graph matching in a pattern recognition context. This algorithm deals with weighted graphs, based on new structural and topological node signatures. Using these signatures, we compute an optimum solution for node-to-node assignment with the Hungarian method and propose a distance formula to compute the distance between weighted graphs. The experiments demonstrate that the newly presented algorithm is well suited to pattern recognition applications. Compared with four well-known methods, our algorithm gives good results for clustering and retrieving images. A sensitivity analysis reveals that the proposed method is also insensitive to weak structural changes.

Graph Embedding Using Constant Shift Embedding

“…Our intention is to show that the proposed graph embedding technique is indeed able to yield embedded vectors that can improve classification results achieved with the original graph representation. We begin by computing the dissimilarities matrix of each data set by means of the graph matching introduced in [10,11] and briefly reviewed in Section 2. Then, since the classification in the graph domain can be performed by only the k-NN classifiers, hence, it is used as reference system in the graph domain.…”

“…Many approaches have been proposed to solve the graph matching problem. In this paper we use a recent technique proposed by Jouili et al in [10,11]. This approach is based on node signatures notion.…”

“…This set of distances can be realized as the pairwise distances among a set of points in an Euclidean space. In our method, we generalize this method by means of graph similarity measure [11,10]. The experimental results have shown that the pro-posed graph embedding technique improves the accuracy achieved in the graph domain by the nearest neighbor classifier.…”

Attributed Graph Matching Using Local Descriptions

“…We have taken inspiration from literature, to use an assignment-based algorithm for graph matching [3,9,18,19] by making use of a new node signature. To compute the distance between graphs, a framework is proposed in this section to extract node signatures and compute distance between these signatures.…”

Comparing Graph Similarity Measures for Graphical Recognition

“…Nevertheless, dealing with graphs suffers, on the one hand from the high complexity of the graph matching problem which is a problem of computing distances between graphs, and on the other hand from the robustness to structural noise which is a problem related to the capability to cope with structural variations and differences in the size of the graph. In order to overcome this problem, several approximate graph matching methods have been proposed [11,15,19,21]. In this paper, our attention is focused on the comparison of different graph similarity measures in the context of document retrieval.…”