Labeled bunch graphs for image analysis

• US 6,563,950 B1
• Filed: 12/21/2001
• Issued: 05/13/2003
• Est. Priority Date: 06/25/1996
• Status: Expired due to Term

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Neven Vision, Inc. (Alphabet Inc.)
• Inventors: von der Malsburg, Christoph, Wiskott, Laurenz
• Primary Examiner(s): Johns, Andrew W.
• Assistant Examiner(s): Mariam, Daniel G.
• Application Number:
  US10/029,691
• Time in Patent Office: 508 Days
• Field of Search: 382/103, 382/118, 382/151, 382/159, 382/181, 382/190, 382/201, 382/209, 382/218, 382/226, 382/240, 382/276, 382/278, 382/282, 382/291, 382/308, 382/180, 345/419, 345/473, 345/619, 345/646, 348/77, 348/169
• US Class Current: 382/209
• CPC Class Codes: G06V 10/426 Graphical representations
• CPC Class Codes: G06V 10/449 Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
• CPC Class Codes: G06V 30/1988 Graph matching graphical image representation G06V30/18181
• 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 process for image analysis which includes selecting a number M of images, forming a model graph from each of the number of images, such that each model has a number N of nodes, assembling the model graphs into a gallery, and mapping the gallery of model graphs into an associated bunch graph by using average distance vectors Δ_ij for the model graphs as edge vectors in the associated bunch graph. A number M of jets is associated with each node of the associated bunch graph, and at least one jet is labeled with an attribute characteristic of one of the number of images. An elastic graph matching procedure is performed wherein the graph similarity function is replaced by a bunch-similarity function.

Claims

1. A method for analysis of a feature in an image having an array of pixel positions, comprising the steps of:

selecting a pixel position in the image associated with the feature;

extracting a jet at the pixel position using wavelet transformations, the jet being represented by a predetermined number of coefficients; and

comparing the extracted jet with a plurality of reference jets, each reference jet being associated through a human assisted process with an attribute label for a particular feature type, to select an attribute label for the feature.

2. A method for analysis of a feature in an image as defined in claim 1, wherein the wavelet transformations use Gabor wavelets.

3. A method for analysis of a feature in an image as defined in claim 1, wherein each coefficient of a jet is generated based on a Gabor wavelet having a particular orientation and frequency.

4. A method for analysis of a feature in an image as defined in claim 1, wherein the predetermined number of coefficients representing the jet is 40.

5. A method for analysis of a feature in an image as defined in claim 1, wherein the coefficients are of complex numbers.

6. A method for analysis of a feature in an image as defined in claim 1, wherein the feature in the image is a facial feature.

7. A method for analysis of features in an image having an array of pixel positions, comprising the steps of:

selecting pixel positions in the image corresponding to nodes of a graph, each node for analyzing a local feature in the image;

extracting a jet at each selected pixel position using wavelet transformations, each jet being represented by a predetermined number of coefficients; and

comparing each of the extracted jets with a plurality of reference jets, each reference jet being associated through a human assisted process with an attribute label for a particular local feature type, to select an attribute label for each local feature associated with a node of the graph.

8. A method for analysis of features in an image as defined in claim 7, wherein the wavelet transformations use Gabor wavelets.

9. A method for analysis of features in an image as defined in claim 7, wherein each coefficient of a jet is generated based on a Gabor wavelet having a particular orientation and frequency.

10. A method for analysis of features in an image as defined in claim 7, wherein the predetermined number of coefficients representing a jet is 40.

11. A method for analysis of features in an image as defined in claim 7, wherein the coefficients are complex numbers.

12. A method for analysis of features in an image as defined in claim 7, wherein the features in the image are facial features.

13. A method for analysis of features in an image as defined in claim 7, further comprising determining an image feature type based on a vote of the attribute labels selected for the local features.