Difference between revisions of "Publications/xu.16.pami"

From LRDE

Line 5: Line 5:
 
| title = Hierarchical Segmentation Using Tree-Based Shape Spaces
 
| title = Hierarchical Segmentation Using Tree-Based Shape Spaces
 
| journal = IEEE Transactions on Pattern Analysis and Machine Intelligence
 
| journal = IEEE Transactions on Pattern Analysis and Machine Intelligence
| volume = PP
+
| volume = 39
| number = 99
+
| number = 3
| pages = 1 to 14
+
| pages = 457 to 469
 
| lrdeprojects = Image
 
| lrdeprojects = Image
 
| abstract = Current trends in image segmentation are to compute a hierarchy of image segmentations from fine to coarse. A classical approach to obtain a single meaningful image partition from a given hierarchy is to cut it in an optimal way, following the seminal approach of the scale-set theory. While interesting in many cases, the resulting segmentation, being a non-horizontal cut, is limited by the structure of the hierarchy. In this paper, we propose a novel approach that acts by transforming an input hierarchy into a new saliency map. It relies on the notion of shape space: a graph representation of a set of regions extracted from the image. Each region is characterized with an attribute describing it. We weigh the boundaries of a subset of meaningful regions (local minima) in the shape space by extinction values based on the attribute. This extinction-based saliency map represents a new hierarchy of segmentations highlighting regions having some specific characteristics. Each threshold of this map represents a segmentation which is generally different from any cut of the original hierarchy. This new approach thus enlarges the set of possible partition results that can be extracted from a given hierarchy. Qualitative and quantitative illustrations demonstrate the usefulness of the proposed method.
 
| abstract = Current trends in image segmentation are to compute a hierarchy of image segmentations from fine to coarse. A classical approach to obtain a single meaningful image partition from a given hierarchy is to cut it in an optimal way, following the seminal approach of the scale-set theory. While interesting in many cases, the resulting segmentation, being a non-horizontal cut, is limited by the structure of the hierarchy. In this paper, we propose a novel approach that acts by transforming an input hierarchy into a new saliency map. It relies on the notion of shape space: a graph representation of a set of regions extracted from the image. Each region is characterized with an attribute describing it. We weigh the boundaries of a subset of meaningful regions (local minima) in the shape space by extinction values based on the attribute. This extinction-based saliency map represents a new hierarchy of segmentations highlighting regions having some specific characteristics. Each threshold of this map represents a segmentation which is generally different from any cut of the original hierarchy. This new approach thus enlarges the set of possible partition results that can be extracted from a given hierarchy. Qualitative and quantitative illustrations demonstrate the usefulness of the proposed method.
| note = To appear
 
 
| lrdepaper = http://www.lrde.epita.fr/dload/papers/xu.16.pami.pdf
 
| lrdepaper = http://www.lrde.epita.fr/dload/papers/xu.16.pami.pdf
 
| lrdekeywords = Image
 
| lrdekeywords = Image
Line 23: Line 22:
 
journal = <nowiki>{</nowiki>IEEE Transactions on Pattern Analysis and Machine
 
journal = <nowiki>{</nowiki>IEEE Transactions on Pattern Analysis and Machine
 
Intelligence<nowiki>}</nowiki>,
 
Intelligence<nowiki>}</nowiki>,
year = <nowiki>{</nowiki>2016<nowiki>}</nowiki>,
+
year = <nowiki>{</nowiki>2017<nowiki>}</nowiki>,
volume = <nowiki>{</nowiki>PP<nowiki>}</nowiki>,
+
volume = <nowiki>{</nowiki>39<nowiki>}</nowiki>,
number = <nowiki>{</nowiki>99<nowiki>}</nowiki>,
+
number = <nowiki>{</nowiki>3<nowiki>}</nowiki>,
pages = <nowiki>{</nowiki>1-14<nowiki>}</nowiki>,
+
pages = <nowiki>{</nowiki>457--469<nowiki>}</nowiki>,
  +
month = apr,
 
abstract = <nowiki>{</nowiki>Current trends in image segmentation are to compute a
 
abstract = <nowiki>{</nowiki>Current trends in image segmentation are to compute a
 
hierarchy of image segmentations from fine to coarse. A
 
hierarchy of image segmentations from fine to coarse. A
Line 50: Line 50:
 
from a given hierarchy. Qualitative and quantitative
 
from a given hierarchy. Qualitative and quantitative
 
illustrations demonstrate the usefulness of the proposed
 
illustrations demonstrate the usefulness of the proposed
method.<nowiki>}</nowiki>,
+
method.<nowiki>}</nowiki>
note = <nowiki>{</nowiki>To appear<nowiki>}</nowiki>
 
 
<nowiki>}</nowiki>
 
<nowiki>}</nowiki>
   

Revision as of 15:08, 9 February 2017

Abstract

Current trends in image segmentation are to compute a hierarchy of image segmentations from fine to coarse. A classical approach to obtain a single meaningful image partition from a given hierarchy is to cut it in an optimal way, following the seminal approach of the scale-set theory. While interesting in many cases, the resulting segmentation, being a non-horizontal cut, is limited by the structure of the hierarchy. In this paper, we propose a novel approach that acts by transforming an input hierarchy into a new saliency map. It relies on the notion of shape space: a graph representation of a set of regions extracted from the image. Each region is characterized with an attribute describing it. We weigh the boundaries of a subset of meaningful regions (local minima) in the shape space by extinction values based on the attribute. This extinction-based saliency map represents a new hierarchy of segmentations highlighting regions having some specific characteristics. Each threshold of this map represents a segmentation which is generally different from any cut of the original hierarchy. This new approach thus enlarges the set of possible partition results that can be extracted from a given hierarchy. Qualitative and quantitative illustrations demonstrate the usefulness of the proposed method.

Documents

Bibtex (lrde.bib)

@Article{	  xu.16.pami,
  author	= {Yongchao Xu and Edwin Carlinet and Thierry G\'eraud and
		  Laurent Najman},
  title		= {Hierarchical Segmentation Using Tree-Based Shape Spaces},
  journal	= {IEEE Transactions on Pattern Analysis and Machine
		  Intelligence},
  year		= {2017},
  volume	= {39},
  number	= {3},
  pages		= {457--469},
  month		= apr,
  abstract	= {Current trends in image segmentation are to compute a
		  hierarchy of image segmentations from fine to coarse. A
		  classical approach to obtain a single meaningful image
		  partition from a given hierarchy is to cut it in an optimal
		  way, following the seminal approach of the scale-set
		  theory. While interesting in many cases, the resulting
		  segmentation, being a non-horizontal cut, is limited by the
		  structure of the hierarchy. In this paper, we propose a
		  novel approach that acts by transforming an input hierarchy
		  into a new saliency map. It relies on the notion of shape
		  space: a graph representation of a set of regions extracted
		  from the image. Each region is characterized with an
		  attribute describing it. We weigh the boundaries of a
		  subset of meaningful regions (local minima) in the shape
		  space by extinction values based on the attribute. This
		  extinction-based saliency map represents a new hierarchy of
		  segmentations highlighting regions having some specific
		  characteristics. Each threshold of this map represents a
		  segmentation which is generally different from any cut of
		  the original hierarchy. This new approach thus enlarges the
		  set of possible partition results that can be extracted
		  from a given hierarchy. Qualitative and quantitative
		  illustrations demonstrate the usefulness of the proposed
		  method.}
}