Reversed Sketch: A scalable and comparable shape representation

作者：

Highlights：

• We proposed a polygonal shape representation which can simplify the shape features for better processing and comparison.

• We proposed a polygonal contour detection algorithm which can quickly extract contour from a binary or segment image.

• We proposed a comparable metric which can measure the similarity between polygons and the corresponding comparison algorithm. Made a great improvement on the existing metric, our metric and algorithm can be applied to concave or even spiral polygons.

• Our shape representation can be used to build hierarchical index on large image database for content-based image retrieval.

摘要

•We proposed a polygonal shape representation which can simplify the shape features for better processing and comparison.•We proposed a polygonal contour detection algorithm which can quickly extract contour from a binary or segment image.•We proposed a comparable metric which can measure the similarity between polygons and the corresponding comparison algorithm. Made a great improvement on the existing metric, our metric and algorithm can be applied to concave or even spiral polygons.•Our shape representation can be used to build hierarchical index on large image database for content-based image retrieval.

论文关键词：Shape representation,Image matching,Contour detection,Polygon evolution,Content-based image retrieval

论文评审过程：Received 9 July 2016, Revised 4 December 2017, Accepted 3 March 2018, Available online 5 March 2018, Version of Record 26 March 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.03.001