Method and apparatus for representing and searching for an object using shape

• US 7,877,414 B2
• Filed: 10/31/2007
• Issued: 01/25/2011
• Est. Priority Date: 04/29/1999
• Status: Expired due to Fees

• Current Assignee: Rakuten Group, Inc.
• Original Assignee: Mitsubishi Denki Kabushiki Kaisha (Mitsubishi Electric Corporation)
• Inventors: Bober, Miroslaw
• Primary Examiner(s): Woo; Isaac M
• Application Number:
  US11/980,575
• Publication Number:
  US 20080065615A1
• Time in Patent Office: 1,182 Days
• Field of Search: 707600-831
• US Class Current: 707/791
• CPC Class Codes: G06F 16/5854 using shape and object relationship
• CPC Class Codes: G06F 16/7837 using objects detected or recognised in the video content
• CPC Class Codes: G06F 16/7854 using shape G06F16/7837 takes precedence
• CPC Class Codes: G06V 10/46 Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features colour feature extraction G06V10/56
• CPC Class Codes: G06V 10/752 Contour matching
• CPC Class Codes: G06V 20/64 Three-dimensional objects
• CPC Class Codes: Y10S 707/99931 Database or file accessing
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

A method is disclosed for representing a sequence of images constituting a moving image by processing signals corresponding to the image. An object appearing in one image is identified in the sequence in a first perspective view, and the same object appearing in another image is identified in the sequence in a second perspective view. A view descriptor of the outline of the object in the first perspective view is derived and at least one additional view descriptor of the outline of the object in another perspective view is also derived. The two or more view descriptors are associated to form a descriptor which is a single indexable entity for the sequence of images.

Claims

1. A method of representing an object appearing in an image, wherein the object appears in the image with a first two-dimensional outline in a first perspective view, by processing signals corresponding to the image, the method comprising:

extracting the first two-dimensional outline of the object in the image;

deriving, using a curvature scale space representation, a view descriptor of the first two-dimensional outline of the object in the first perspective view;

identifying the object;

obtaining at least one additional outline of the object in a perspective view different from said first perspective view

deriving, using a curvature scale space representation, at least one additional view descriptor of an outline of the object in a perspective view different from said first perspective view; and

associating the two or more view descriptors to form an object descriptor.

2. A method as claimed in claim 1 wherein the image is part of a sequence of images, and at least one additional view descriptor corresponds to a view of the object appearing elsewhere in the image sequence.

3. A method as claimed in claim 1 wherein the image is from a video.

4. A method as claimed in claim 1 further comprising deriving a descriptor of the object which is related to the shape and/or size of the object and which is independent of the view of the object in the image.

9. An apparatus comprising:

a descriptor database storing descriptors of an object or parts of an object appearing in images; and

a processor configured to perform the method of claim 1 or claim 5.

10. A computer-readable storage medium storing computer-executable process steps for implementing a method as claimed in claim 1 or claim 5.

11. A computer system comprising:

a descriptor database storing descriptors of objects or parts of objects appearing in images; and

a control unit configured to perform the method of claim 1 or claim 5.

5. A method of searching for an object in an image by processing signals corresponding to the image, the method comprising:

inputting a query in the form of at least one two-dimensional outline of an object;

deriving a descriptor of the query object;

comparing said query descriptor with stored object descriptors for objects in images derived in accordance with a method comprising the steps of

deriving a view descriptor of the first outline of the object and deriving at least one additional view descriptor of the outline of the object in a different view; and

associating the two or more view descriptors to form an object descriptor, wherein the view descriptors are derived using a curvature scale space representation; and

selecting and displaying at least one result corresponding to an image containing an object for which the comparison between the respective descriptor and the query descriptor indicates a degree of similarity between the query and said object.

6. A method as claimed in claim 5 wherein a query is input in the form of two or more two-dimensional outline of an object, and wherein a query view descriptor is derived for each said outline, and wherein the step of comparing comprises comparing each said query view descriptor with each view descriptor in each stored object descriptor to derive a plurality of view-similarity values.

7. A method as claimed in claim 5 wherein the view-similarity values are analyzed to derive object similarity values.

8. A method as claimed in claim 5 wherein a least some of the object descriptors include view-independent descriptors and wherein the method comprises inputting a view-independent query value and the step of comparing compares the query value with the view-independent descriptors for the stored object descriptors.