Region and edge-adaptive sampling and boundary completion for segmentation
- Los Alamos National Laboratory
Edge detection produces a set of points that are likely to lie on discontinuities between objects within an image. We consider faces of the Gabriel graph of these points, a sub-graph of the Delaunay triangulation. Features are extracted by merging these faces using size, shape and color cues. We measure regional properties of faces using a novel shape-dependant sampling method that overcomes undesirable sampling bias of the Delaunay triangles. Instead, sampling is biased so as to smooth regional statistics within the detected object boundaries, and this smoothing adapts to local geometric features of the shape such as curvature, thickness and straightness.
- Research Organization:
- Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- AC52-06NA25396
- OSTI ID:
- 1011479
- Report Number(s):
- LA-UR-10-03018; LA-UR-10-3018; TRN: US201109%%455
- Resource Relation:
- Journal Volume: 6454; Conference: British Machine Vision Conference 2010 ; August 30, 2010 ; Aberystwyth, Wales, UK
- Country of Publication:
- United States
- Language:
- English
Similar Records
Patch-based image segmentation of satellite imagery using minimum spanning tree construction
Proximity graphs based multi-scale image segmentation
Image segmentation by hierarchial agglomeration of polygons using ecological statistics
Conference
·
Fri Jan 01 00:00:00 EST 2010
·
OSTI ID:1011479
Proximity graphs based multi-scale image segmentation
Conference
·
Tue Jan 01 00:00:00 EST 2008
·
OSTI ID:1011479
Image segmentation by hierarchial agglomeration of polygons using ecological statistics
Patent
·
Tue Apr 23 00:00:00 EDT 2013
·
OSTI ID:1011479