Method and system for fuzzy clustering of images

• US 7,796,821 B2
• Filed: 11/11/2008
• Issued: 09/14/2010
• Est. Priority Date: 03/28/2001
• Status: Expired due to Fees

• Current Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Tarokh, Vahid, Jafarkhani, Hamid
• Primary Examiner(s): Carter; Aaron W
• Application Number:
  US12/268,960
• Publication Number:
  US 20090074305A1
• Time in Patent Office: 672 Days
• Field of Search: 382115-128, 382/181, 382/224, 382/225, 382155-161, 382226-228, 382232-253, 706 1- 9, 706/52, 706/900, 704/245
• US Class Current: 382/224
• CPC Class Codes: G06F 18/2323 based on graph theory, e.g. minimum spanning trees [MST] or graph cuts

Abstract

An approach to clustering a set of images based on similarity measures employs a fuzzy clustering paradigm in which each image is represented by a node in a graph. The graph is ultimately partitioned into subgraphs, each of which represent true clusters among which the various images are distributed. The partitioning is performed in a series of stages by identifying one true cluster at each stage, and removing the nodes belonging to each identified true cluster from further consideration so that the remaining, unclustered nodes may then be grouped. At the beginning of each such stage, the nodes that remain to be clustered are treated as all belonging to a single candidate cluster. Nodes are removed from this single candidate cluster in accordance with similarity and connectivity criteria, to arrive at a true cluster. The member nodes of this true cluster are then removed from further consideration, prior to the next stage in the process.

Claims

1. A method for clustering a set of N images into P final clusters, where N and P are integers, comprising:

calculating at least one similarity measure S_i,j between members of each pair of images, where the similarity measure S_i,j represents the similarity measure between an ith image and an jth image with i and j being image indices;

calculating a total connectivity value for each of the images remaining to be clustered, where the total connectivity value for each image being defined as a sum of a function f of the similarity measures associated with that image;

identifying, from among the images remaining to be clustered, a maximum total connectivity value T_max corresponding to an image I_max, where the image I_max belonging to a current cluster which initially includes all images remaining to be clustered;

removing, from the current cluster, at least one image based on at least one of its similarity measure with image I_max and its total connectivity value within the current cluster;

adding, to the current cluster, images having similarity measures that are greater than a first threshold;

calculating, for each image within the current cluster, the total connectivity value based on those images within the current cluster;

removing, from the current cluster, those images having the total connectivity value less than a second threshold; and

repeating the calculating of the total connectivity value based on those images within the current cluster and the removing from the current cluster those images having the total connectivity value less than the second threshold until no further images are removed to thereby establish the current cluster as one of the final clusters.

2. The method of claim 1, further comprising:

removing all images in the current cluster from further consideration; and

repeating the calculating of the total connectivity value for each of the images remaining to be clustered, the identifying, from among the images remaining to be clustered, the maximum total connectivity value T_max corresponding to the image I_max, the removing the at least one image based on the at least one of its similarity measure with image L_max and its total connectivity value within the current cluster, the adding of images having the similarity measures that are greater than the first threshold, the calculating of the total connectivity value based on those images within the current cluster, the removing from the current cluster those images having the total connectivity value less than the second threshold, the repeating of the calculating of the total connectivity value based on those images within the current cluster and the removing from the current cluster those images having the total connectivity value less than the second threshold, and the removing of all images in the current cluster from further consideration, until all N images are assigned to a final cluster.

3. The method of claim 2, further comprising:

applying a non-linear function to each similarity measure S_i,j, prior to calculating the total connectivity value.

4. The method of claim 3, wherein the non-linear function is a transcendental function.

5. The method of claim 4, wherein the transcendental function is a hyperbolic tangent.

6. The method of claim 5, wherein each similarity measure S_i,j is transformed according to a formula tanh(5×

S_i,j), prior to calculating the total connectivity value.

7. The method of claim 2, wherein the function f is an identity function.

8. The method of claim 2, wherein the function f is a nonlinear function.

9. The method of claim 2, wherein in the removing, at least one image is removed from the current cluster based on its similarity measure S_i,j with image I_max, and then at least one other image is removed from the current cluster based on its total connectivity value within the current cluster.

10. The method of claim 2, wherein in the removing, at least one image is removed from the current cluster based on its total connectivity value within the current cluster, and then at least one other image is removed from the current cluster based on its similarity measure S_i,j with image I_max.

11. The method of claim 2, wherein in the removing images having the similarity measure S_i,j with I_max that is less than a third threshold are removed from the current cluster.

12. The method of claim 2, wherein in the removing, a number N1 images having the N1 lowest similarity measures S_i,j with I_max are removed from the current cluster.

13. The method of claim 2, wherein in the removing

images having the total connectivity value which differs from t_max by more than a fourth threshold are removed from the current cluster.

14. The method of claim 13, wherein images having the similarity measure S_i,j with I_max that is less than a third threshold are removed from the current cluster.

15. The method of claim 14, wherein at least one image is removed from the current cluster based on its similarity measure S_i,j with image I_max, and then at least one other image is removed from the current cluster based on its total connectivity value within the current cluster.

16. The method of claim 13, wherein a number N1 images having the N1 lowest similarity measures S_i,j with I_max are removed from the current cluster.

17. The method of claim 16, wherein at least one image is removed from the current cluster based on its similarity measure S_i,j with image I_max, and then at least one other image is removed from the current cluster based on its total connectivity value within the current cluster.

18. An apparatus for clustering a set of N images into P final clusters, where N and P are integers, comprising:

means for calculating at least one similarity measure S_i,j between members of each pair of images, where the similarity measure S_i,j represents the similarity measure between an ith image and an jth image with i and j being image indices;

means for calculating a total connectivity value for each of the images remaining to be clustered, where the total connectivity value for each image being defined as a sum of a function f of the similarity measures associated with that image;

means for identifying, from among the images remaining to be clustered, a maximum total connectivity value t_max corresponding to an image I_max, where the image I_max belonging to a current cluster which initially includes all images remaining to be clustered;

means for removing, from the current cluster, at least one image based on at least one of its similarity measure with image I_max and its total connectivity value within the current cluster;

means for adding, to the current cluster, images having similarity measures that are greater than a first threshold;

means for calculating, for each image within the current cluster, the total connectivity value based on those images within the current cluster; and

means for removing, from the current cluster, those images having the total connectivity value less than a second threshold, wherein the total connectivity value calculating and the image removing are repeated until no further images are removed to thereby establish the current cluster as one of the final clusters.

19. The apparatus of claim 18, further comprising:

means for removing all images in the current cluster from further consideration, wherein one or more of the calculating, the identifying, the removing, and the adding are repeated until all N images are assigned to a final cluster.

20. A non-transitory computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to perform steps of a method for clustering a set of N images into P final clusters, where N and P are integers comprising:

calculating at least one similarity measure S_i,j between members of each pair of images, where the similarity measure S_i,j represents the similarity measure between an ith image and an jth image with i and j being image indices;

calculating a total connectivity value for each of the images remaining to be clustered, where the total connectivity value for each image being defined as a sum of a function f of the similarity measures associated with that image;

identifying, from among the images remaining to be clustered, a maximum total connectivity value T_max corresponding to an image I_max, where the image I_max belonging to a current cluster which initially includes all images remaining to be clustered;

removing, from the current cluster, at least one image based on at least one of its similarity measure with image I_max and its total connectivity value within current cluster;

adding, to the current cluster, images having similarity measures that are greater than a first threshold;

calculating, for each image within the current cluster, the total connectivity value based on those images within current cluster;

removing, from the current cluster, those images having the total connectivity value less than a second threshold; and

repeating the calculating of the total connectivity value based on those images within the current cluster and the removing from the current cluster those images having the total connectivity value less than the second threshold until no further images are removed to thereby establish the current cluster as one of the final clusters.