Image evaluation apparatus that calculates an importance degree of each of a plurality of images

Patent metadata

Field	Value
Publication number	US-8929669-B2
Application number	US-201113391928-A
Country	US
Kind code	B2
Filing date	Jun 8, 2011
Priority date	Jun 23, 2010
Publication date	Jan 6, 2015
Grant date	Jan 6, 2015

Abstract

An image evaluation apparatus is provided, which calculates a characteristic value indicating a state of appearance of objects corresponding to person a and person b appearing in image A and an object corresponding to person b appearing in image B. Subsequently, the image evaluation apparatus specifies person b′ as the photographer of image B and calculates a likelihood degree indicating accuracy of the determination. Further, the image evaluation apparatus calculates an importance degree of each of images A and B according to the characteristic values of person a and person b appearing in image A and image B and the likelihood degree of the photographer of image B.

First claim

The invention claimed is:

1. An image evaluation apparatus that calculates an importance degree of each of a plurality of images including objects, the image evaluation apparatus comprising:
a processor; and
a non-transitory memory storing executable instructions, which when executed, cause the processor to operate as:
a characteristic value calculating unit that calculates, for each of the objects appearing in each of the images, a characteristic value indicating a state of appearance of the object;
a photographer extracting unit that, for each of the images, (i) determines a photographer of the image and (ii) calculates a likelihood degree indicating accuracy of the determination of the photographer; and
an evaluation unit that calculates an importance degree of each of the images according to the characteristic values of the objects in the images and the likelihood degrees of the photographers of the images.

2. The image evaluation apparatus of claim 1 , wherein
the objects included in the images each belong to one of a plurality of object clusters, and
the photographer extracting unit specifies one or more of the object clusters as the photographer of the image.

3. The image evaluation apparatus of claim 2 , wherein
the images are each associated with one of a plurality of event clusters, an event cluster associated with a given image indicating an event that the given image belongs to, and the photographer extracting unit (i) obtains, according to an event cluster associated with the image, other images that belong to the event cluster, (ii) extracts objects from the other images, and (iii) specifies, as the photographer, one or more object clusters corresponding to one or more objects that are not included in the image, among the objects extracted from the other images, and
the photographer extracting unit calculates a likelihood degree for each of the one or more object clusters specified.

4. The image evaluation apparatus of claim 3 , wherein
the photographer extracting unit calculates, for each of the one or more object clusters specified, (i) a scarcity degree indicating how infrequently the object cluster appears in images belonging to the event cluster that the object cluster appears in and (ii) a photographer density degree indicating a frequency at which the object cluster is specified as the photographer in images having a production data and time close to a production date and time of the image, and
the photographer extracting unit calculates a likelihood degree for each of the one or more object clusters according to the scarcity degree and the photographer density degree.

5. The image evaluation apparatus of claim 3 , wherein
when none of the object clusters is specified as the photographer of the image by the photographer extracting unit,
the photographer extracting unit further (i) obtains, according to the event cluster associated with the image, images that belong to event clusters other than the event cluster associated with the image, (ii) extracts objects included in the images that belong to event clusters other than the event cluster associated with the image, and (iii) specifies, as the photographer of the image, an object cluster corresponding to an object that does not appear in the image, among the objects included in the images that belong to event clusters other than the event cluster associated with the image.

6. The image evaluation apparatus of claim 2 , wherein
the images are each associated with one of a plurality of event clusters, an event cluster associated with a given image indicating an event that the given image belongs to, and
the photographer extracting unit (i) obtains, according to the event cluster associated with the image, images that belong to event clusters other than the event cluster associated with the image, (ii) extracts objects included in the images that belong to event clusters other than the event cluster associated with the image, and (iii) specifies, as the photographer of the image, an object cluster corresponding to an object that does not appear in the image, among the objects included in the images that belong to event clusters other than the event cluster associated with the image.

7. The image evaluation apparatus of claim 6 , wherein
when more than one object cluster is specified as the photographer of the image by the photographer extracting unit,
the photographer extracting unit calculates, for each of the more than one object clusters specified, (i) a familiarity degree that indicates familiarity between the object cluster and the object cluster that the object included in the image belongs to, and (ii) a scarcity degree indicating how infrequently the object cluster appears in images belonging to the event cluster that the object cluster appears in, and
the photographer extracting unit calculates, for each of the more than one object clusters specified, a photographer specification degree that is in accordance with the familiarity degree and the scarcity degree and thereby specifies, as the photographer of the image, an object cluster having the greatest photographer specification degree, among the more than one object clusters specified.

8. The image evaluation apparatus of claim 1 , wherein
the photographer extracting unit attempts to obtain information pertaining to the photographer of the image from metadata provided to the image and determines the photographer according to the information obtained.

9. The image evaluation apparatus of claim 1 , wherein
the evaluation unit calculates an importance degree of the photographers of the images in addition to the importance degrees of the images, the calculation of importance degrees being performed such that importance propagates from the importance degrees of the photographers to the importance degrees of the images.

10. The image evaluation apparatus of claim 9 , wherein
the objects included in the images each belong to one of a plurality of object clusters, and
the evaluation unit calculates an importance degree of each of the object clusters in addition to the importance degrees of the images and the importance degrees of the photographers, the calculation of importance degrees being performed such that importance propagates from the importance degrees of the images to the importance degrees of the object clusters.

11. The image evaluation apparatus of claim 1 ,
wherein the processor further operates as:
a creating unit that creates a data structure including
a plurality of image elements respectively indicating the plurality of images,
a plurality of object cluster elements respectively indicating object clusters that the objects belong to, and
a plurality of photographer elements respectively indicating the photographers of the images,

a first link setting unit that sets, for each of the images, a value of a link between the image element corresponding to the image and the photographer element corresponding to the photographer of the image according to the likelihood degree; and
a second link setting unit that sets, for each of the images, a value of a link between an image element corresponding to the image and an object cluster element corresponding to an object cluster that the object belongs to according to the characteristic value, and

the evaluation unit calculates, for each of the images, the importance degree of the image according to the values of the links set with respect to the image element corresponding to the image.

12. The image evaluation apparatus of claim 11 , wherein
the creating unit sets, for each of the images, one of the object cluster elements as a main subject element indicatin

Assignees

• Yabu Hiroshi
• Panasonic Ip Corp America

Inventors

• Yabu Hiroshi

Classifications

• G06V10/7635

  based on graphs, e.g. graph cuts or spectral clustering · CPC title

• G06F18/2323

  based on graph theory, e.g. minimum spanning trees [MST] or graph cuts · CPC title

• G06K9/00677

  Physics · mapped topic

• G06F17/30247 Primary

  Physics · mapped topic

• G06F17/30256

  Physics · mapped topic