Abstract | ||
---|---|---|
In this paper, we describe issues related to the measurement of structural similarity between document images. We define structural similarity, and discuss the benefits of using it as a complement to content similarity for querying document image databases. We present an approach to computing a geometrically invariant structural similarity, and use this measure to search document image databases. Our approach supports both full image matching using query by example (QBE) and sub-image matching using query by sketch (QBS). The similarity measure considers spatial and layout structure, and is computed by aggregating content area overlap measures with respect to their underlying column structures. These techniques are tested within the Intelligent Document Image Retrieval (IDIR) System, and results demonstrating effectiveness and efficiency of structure queries with respect to human relevance judgments are presented. |
Year | Venue | Keywords |
---|---|---|
2006 | IPCV | index terms—document image understanding,indexing and retrieval of document images,similarity,document layout structure,indexing terms,query by example,image retrieval,structural similarity |
Field | DocType | Citations |
Automatic image annotation,Similarity measure,Pattern recognition,Information retrieval,Image texture,Computer science,Image retrieval,Query by Example,Invariant (mathematics),Artificial intelligence,Intelligent document,Visual Word | Conference | 11 |
PageRank | References | Authors |
0.72 | 9 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Christian Shin | 1 | 78 | 7.26 |
David Doermann | 2 | 4313 | 312.70 |