Segmenting an image
AssigneeHewlett-Packard Development Company, L.P.
InventorDaniel Freedman
Life of this patent
Estimating the term…
Abstract
A technique includes representing an image to be segmented using a graph including vertices and edges, where a subset of the vertices is marked as belonging to a plurality of objects. The technique includes assigning weights to the edges based at least in part on color probability densities associated with the objects derived from the subset of vertices marked as belonging to the plurality of objects; and segmenting the image, including identifying boundaries of the objects based at least in part on the weights and the graph.
Claims20
A method comprising:
- representing an image to be segmented using a graph comprising vertices and edges, a subset of the vertices being marked as belonging to a plurality of objects;
- using a processor-based machine to assign weights to the edges based at least in part on color probability densities associated with the objects derived from the subset of vertices marked as belonging to the plurality of objects;
- and
- segmenting the image, comprising identifying boundaries of the objects based at least in part on the weights and the graph.
A system comprising:
- a memory to store data indicative of an image to be segmented;
- and
- at least one machine-based processing engine adapted to:
- process the data to construct a graph comprising vertices and edges, a subset of the vertices being marked as belonging to a plurality of objects;
- and
- assign weights to the edges based at least in part on color probability densities associated with the objects derived from the subset of vertices marked as belonging to the plurality of objects;
- and
- segment the image, comprising identifying boundaries of the objects based at least in part on the weights and the graph.
An article comprising a storage readable by at least one processor storing instructions that when executed by said at least one processor causes said at least one processor to:
- represent an image to be segmented using a graph comprising vertices and edges, a subset of the vertices being marked as belonging to a plurality of objects;
- assign weights to the edges based at least in part on color probability densities associated with the objects derived from the subset of vertices marked as belonging to the plurality of objects;
- and
- segment the image, comprising identifying boundaries of the objects based at least in part on the weights and the graph.
Dependencies are inferred from the claim text.