Method and apparatus for generating superpixel clusters

Patent metadata

Field	Value
Publication number	US-9971951-B2
Application number	US-201415026706-A
Country	US
Kind code	B2
Filing date	Sep 22, 2014
Priority date	Oct 2, 2013
Publication date	May 15, 2018
Grant date	May 15, 2018

Abstract

Method and apparatus for generating superpixel clusters A method and an apparatus ( 20 ) for generating a superpixel cluster for an image are described. A clustering unit ( 23 ) generates ( 10 ) an initial superpixel cluster based on an initial cluster center. The clustering unit ( 23 ) further generates ( 12 ) one or more further superpixel clusters based on one or more further cluster centers. A determining unit ( 24 ) determines ( 11 ) the one or more further cluster centers from the initial superpixel cluster. A merger ( 25 ) then merges ( 13 ) the initial superpixel cluster and the one or more further superpixel clusters to obtain the superpixel cluster for the image.

First claim

The invention claimed is:

1. A method for generating a superpixel cluster for an image, the method comprising:
generating an initial cluster of superpixels from an initial cluster center;
determining, within said initial cluster, one or more further cluster centers different from said initial cluster center;
generating one or more further clusters of superpixels from the one or more further cluster centers; and
grouping the superpixels of the initial cluster with the superpixels of the one or more further superpixel clusters to obtain the superpixel cluster for the image.

2. The method according to claim 1 , wherein the initial cluster and the one or more further clusters are generated using a centralistic cluster building approach.

3. The method according to claim 2 , wherein the initial cluster is generated by comparing connectivity attributes of an initial superpixel comprising said initial cluster center with connectivity attributes of superpixels adjacent to said initial superpixel and the one or more further clusters are generated by comparing connectivity attributes of further superpixels comprising said further cluster centers with connectivity attributes of superpixels adjacent to said further superpixels.

4. The method according to claim 3 , wherein the connectivity attributes are color histograms.

5. The method according to claim 3 , wherein a decision on the cluster inclusion or exclusion of a superpixel is taken by means of a distance measure calculated between the connectivity attributes associated with said superpixel and the connectivity attributes associated with the superpixel comprising the center of said cluster.

6. The method according to claim 5 , wherein a superpixel is excluded from a cluster if the distance measure exceeds a threshold.

7. An apparatus configured to generate a superpixel cluster for an image, the apparatus comprising a processor configured to:
generate an initial cluster of superpixels from an initial cluster center;
determine, within said initial cluster, one or more further cluster centers different from said initial cluster center;
generate one or more further clusters of superpixels from the one or more further cluster centers; and
group the superpixels of the initial cluster with the superpixels of the one or more further superpixel clusters to obtain the superpixel cluster for the image.

8. The apparatus according to claim 7 , wherein processor is configured to generate the initial cluster and the one or more further clusters using a centralistic cluster building approach.

9. The apparatus according to claim 8 , wherein the processor is further configured to generate the initial cluster by comparing connectivity attributes of an initial superpixel comprising said initial cluster center with connectivity attributes of superpixels adjacent to said initial superpixel and to generate the one or more further clusters by comparing connectivity attributes of further superpixels comprising said further cluster centers with connectivity attributes of superpixels adjacent to said further superpixels.

10. The apparatus according to claim 9 , wherein the connectivity attributes are color histograms.

11. The apparatus according to claim 9 , wherein said processor is configured to take a decision on the cluster inclusion or exclusion of a superpixel by means of a distance measure calculated between the connectivity attributes associated with said superpixel and the connectivity attributes associated with the superpixel comprising the center of said cluster.

12. The apparatus according to claim 11 , wherein said processor is configured to exclude a superpixel from a cluster if the distance measure exceeds a threshold.

13. A non-transitory computer readable storage medium having stored therein instructions enabling generating a superpixel cluster for an image, which, when executed by a computer, cause the computer to:
generate an initial cluster of superpixels from an initial cluster center;
determine, within said initial cluster, one or more further cluster centers different from the initial cluster center;
generate one or more further clusters of superpixels from the one or more further cluster centers; and
group the superpixels of the initial cluster with the superpixels of the one or more further superpixel clusters to obtain the superpixel cluster for the image.

14. The non-transitory computer readable storage medium according to claim 13 , wherein the instructions cause the computer to generate the initial superpixel cluster and the one or more further superpixel clusters using a centralistic cluster building approach.

15. The non-transitory computer readable storage medium according to claim 14 , wherein the instructions cause the computer to generate the initial cluster by comparing connectivity attributes of an initial superpixel comprising said initial cluster center with connectivity attributes of superpixels adjacent to said initial superpixel and to generate the one or more further clusters by comparing connectivity attributes of further superpixels comprising said further cluster centers with connectivity attributes of adjacent superpixels adjacent to said further superpixels.

16. The non-transitory computer readable storage medium according to claim 15 , wherein the connectivity attributes are color histograms.

17. The non-transitory computer readable storage medium according to claim 15 , wherein the instructions cause the computer to take a decision on the cluster inclusion or exclusion of a superpixel by means of a distance measure calculated between the connectivity attributes associated with said superpixel and the connectivity attributes associated with the superpixel comprising the center of said cluster.

18. The non-transitory computer readable storage medium according to claim 17 , wherein the instructions cause the computer to exclude a superpixel from a cluster if the distance measure exceeds a threshold.

Assignees

• Thomson Licensing

Inventors

• Gandolph Dirk
• Putzke-Roeming Wolfram
• Schewzow Andrej

Classifications

• G06T7/11 Primary

  Region-based segmentation · CPC title

• G06K9/46 Primary

  Physics · mapped topic

• G06T7/187

  involving region growing; involving region merging; involving connected component labelling · CPC title