Near-duplicate document image matching: A graphical perspective

作者:

Highlights:

• We propose a near-duplicate document image matching approach.

• Document images are represented by graphs.

• The nodes correspond to the objects in the images.

• The edges capture the relations among the objects.

• A multi-granularity object tree is built to settle the instability of object segmentation.

摘要

Highlights•We propose a near-duplicate document image matching approach.•Document images are represented by graphs.•The nodes correspond to the objects in the images.•The edges capture the relations among the objects.•A multi-granularity object tree is built to settle the instability of object segmentation.

论文关键词:Document images,Near-duplicate documents,Document image matching,Graph representation,Multi-granularity object tree,Graph matching

论文评审过程:Received 18 June 2013, Revised 22 September 2013, Accepted 4 November 2013, Available online 18 November 2013.

论文官网地址:https://doi.org/10.1016/j.patcog.2013.11.006