Image congealing via efficient feature selection

• US 9,477,905 B2
• Filed: 07/13/2015
• Issued: 10/25/2016
• Est. Priority Date: 01/09/2012
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Liu, Xiaoming, Tu, Peter Henry, Xue, Ya
• Primary Examiner(s): Tabatabai, Abolfazl
• Application Number:
  US14/798,160
• Publication Number:
  US 20150324663A1
• Time in Patent Office: 470 Days
• Field of Search: 382/164, 382/171, 382/173, 382/179, 382/215, 711/206, 711/208, 711/209, 348/267
• US Class Current: 1/1
• CPC Class Codes: G06F 18/2113 by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
• CPC Class Codes: G06F 18/2178 based on feedback of a supervisor
• CPC Class Codes: G06F 18/2323 based on graph theory, e.g. minimum spanning trees [MST] or graph cuts
• CPC Class Codes: G06V 10/754 involving a deformation of the sample pattern or of the reference pattern; Elastic matching
• CPC Class Codes: G06V 10/7635 based on graphs, e.g. graph cuts or spectral clustering
• CPC Class Codes: G06V 10/771 Feature selection, e.g. selecting representative features from a multi-dimensional feature space
• CPC Class Codes: G06V 10/7784 based on feedback from supervisors
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships

Abstract

A novel technique for unsupervised feature selection is disclosed. The disclosed methods include automatically selecting a subset of a feature of an image. Additionally, the selection of the subset of features may be incorporated with a congealing algorithm, such as a least-square-based congealing algorithm. By selecting a subset of the feature representation of an image, redundant and/or irrelevant features may be reduced or removed, and the efficiency and accuracy of least-square-based congealing may be improved.

Claims

1. A method, comprising:

incorporating an unsupervised feature selection algorithm with an image congealing algorithm;

executing, via a processor, the unsupervised feature selection algorithm to determine representative features of an image; and

executing, via a processor, the image congealing algorithm to estimate warping parameters for the image using the representative features.

2. The method of claim 1, wherein executing the unsupervised feature selection algorithm comprises constructing a graph having features of the image as vertices.

3. The method of claim 2, comprising predetermining a number of the features.

4. The method of claim 2, wherein executing the unsupervised feature selection algorithm comprises determining a connectivity between the vertices using a maximum information compression index.

5. The method of claim 4, wherein executing the unsupervised feature selection algorithm comprises partitioning the graph into two or more subsets of features using a power iteration clustering algorithm.

6. The method of claim 5, wherein executing the unsupervised feature selection algorithm comprises selecting representative features from each subset of features.

7. The method of claim 5, comprising adding a perturbation to an initial vector of the power iteration clustering algorithm.

8. The method of claim 4, comprising reducing computational cost of the power iteration clustering algorithm by using a fast k-means algorithm.

9. The method of claim 1, wherein executing the image congealing algorithm to estimate warping parameters for the image using the representative features comprises computing warping parameter updates for the image and updating the warping parameters of the image to calculate current warping parameters.

10. The method of claim 1, comprising initially estimating the warping parameters before executing the image congealing algorithm to estimate warping parameters for the image using the representative features.

11. A method, comprising:

executing, via a processor, an unsupervised feature selection algorithm, comprising;

inputting a data matrix, where rows of the data matrix comprise instances of an image, and columns of the data matrix comprise features of the image;

calculating a similarity between each pair of the features;

generating a graph with the features as vertices of the graph;

clustering the features of the graph into a plurality of groups; and

for each of the plurality of groups, selecting a representative feature that is closest to a center of the respective group.

12. The method of claim 11, comprising inputting a desired number of the features with the data matrix.

13. The method of claim 11, comprising incorporating the unsupervised feature selection algorithm with an unsupervised congealing framework.

14. The method of claim 13, wherein the unsupervised congealing framework comprises a least-square-based image congealing algorithm.

15. The method of claim 11, wherein the unsupervised feature selection algorithm comprises a filter-type unsupervised feature selection algorithm.

16. The method of claim 11, wherein clustering the features of the graph into the plurality of groups comprises applying a power iteration clustering algorithm.

17. The method of claim 11, wherein calculating the similarity between each pair of the features comprises applying a heuristic algorithm or a spectral clustering algorithm.

18. A method, comprising:

executing, via a processor, an unsupervised feature selection algorithm, comprising;

constructing a graph having features of an image as vertices;

partitioning the graph into two or more subsets of features; and

selecting a representative feature from each of the two or more subsets; and

executing an image congealing algorithm to estimate warping parameters for the image using the representative features.

19. The method of claim 18, wherein a number of the representative features is less than 3% of a total number of the features.

20. The method of claim 18, comprising integrating the unsupervised feature selection algorithm with the image congealing algorithm.