Pattern recognition system and method using Gabor functions

• US 9,058,517 B1
• Filed: 04/16/2014
• Issued: 06/16/2015
• Est. Priority Date: 04/16/2014
• Status: Active Grant

• Current Assignee: I.R.I.S. (Fidelity National Information Services Incorporated)
• Original Assignee: I.R.I.S. (Fidelity National Information Services Incorporated)
• Inventors: Dauw, Michel, De Muelenaere, Pierre, Dupont, Olivier, Collet, Frederic, Hautot, Jordi, Hensges, Gunter
• Primary Examiner(s): Rush, Eric
• Application Number:
  US14/254,039
• Time in Patent Office: 426 Days
• Field of Search: 382100-102, 382/105, 382112-114, 382135-138, 382/160, 382/161, 382/181, 382/182, 382185-187, 382/190, 382/198, 382/200, 382/209, 382/224, 382/279, 382/321
• US Class Current: 1/1
• CPC Class Codes: G06F 16/583 using metadata automatically derived from the content
• CPC Class Codes: G06F 18/2136 based on sparsity criteria, e.g. with an overcomplete basis
• CPC Class Codes: G06F 2218/02 Preprocessing
• CPC Class Codes: G06F 2218/08 Feature extraction
• CPC Class Codes: G06F 2218/10 by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
• CPC Class Codes: G06F 2218/12 Classification; Matching
• CPC Class Codes: G06F 40/279 Recognition of textual entities
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/18038 Biologically-inspired filters, e.g. difference of Gaussians [DoG], Gabor filters
• CPC Class Codes: G06V 30/19127 Extracting features by transforming the feature space, e.g. multidimensional scaling; Mappings, e.g. subspace methods

Abstract

A pattern recognition system and method which generates a feature vector by multiplying an image vector with a sparse matrix. The sparse matrix is generated from a Gabor function which is a sinusoidal wave multiplied by a Gaussian function. The Gabor function is a function of a set of parameters including a parameter related to the direction of the sinusoidal wave, a parameter related to a center of the Gabor function, and a parameter related to a wavelength of the sinusoidal wave. The wavelength takes at least two values, with a first wavelength value lower than or substantially equal to the distance between two adjacent centers of the Gabor function, and the first wavelength value is lower than a second wavelength value and higher than or substantially equal to half the second wavelength value.

Claims

1. A method for identifying a pattern in an input image, comprising the steps of

a) normalizing the input image to a normalized matrix representing a normalized image,

b) generating an image vector from the normalized matrix,

c) multiplying the image vector with a sparse matrix using a matrix vector multiplication to generate a feature vector wherein the sparse matrix is generated from a Gabor function which is a sinusoidal wave multiplied by a Gaussian function and wherein the Gabor function is a function of at least one variable indicating a position in the normalized matrix and of a set of parameters including a parameter related to the direction of the sinusoidal wave, a parameter related to a centre of the Gabor function, and a parameter related to a wavelength of the sinusoidal wave,

d) creating with the feature vector a density of probability for a predetermined list of models,

e) selecting the model with the highest density of probability as the best model, and

f) classifying the best model as the pattern of the input image,

wherein there are at least two centres of the Gabor function, and

wherein the wavelength takes at least two values, with a first wavelength value lower than or substantially equal to the distance between two adjacent centres of the Gabor function, and the first wavelength value is lower than a second wavelength value and higher than or substantially equal to half the second wavelength value

wherein the models are characterized by a covariance matrix and by an average vector,

wherein the density of probability is calculated by the formula

2. A method according to claim 1, wherein all non-diagonal elements of the covariance matrix are zero.

3. A method according to claim 1, wherein the parameter related to the direction of the sinusoidal wave is an angle.

4. A method according to claim 1, with at least two parameters related to a centre of the Gabor function,

wherein the parameters related to a centre of the Gabor function are such that the centres of the Gabor function are regularly spaced.

5. A method according to claim 1, wherein the feature vector is approximated.

6. A computer program product comprising a non-transitory computer readable medium having control logic stored therein for causing a computing device to identify a pattern in an input image, the control logic comprising:

a) first computer readable program code means for normalizing the input image to a normalized matrix representing a normalized image,

b) second computer readable program code means for generating an image vector from the normalized matrix,

c) third computer readable program code means for multiplying the image vector with a sparse matrix using a matrix vector multiplication to generate a feature vector wherein the sparse matrix (303) is generated from a Gabor function which is a sinusoidal wave multiplied by a Gaussian function and wherein the Gabor function is a function of at least one variable indicating a position in the normalized matrix and of a set of parameters including a parameter related to the direction of the sinusoidal wave, a parameter related to a centre of the Gabor function, and a parameter related to a wavelength of the sinusoidal wave,

d) fourth computer readable program code means for creating with the feature vector a density of probability for a predetermined list of models,

e) fifth computer readable program code means for selecting the model with the highest density of probability as the best model, and

f) sixth computer readable program code means for classifying the best model as the pattern of the input image,

wherein there are at least two centres of the Gabor function, and

wherein the wavelength takes at least two values, with a first wavelength value lower than or substantially equal to the distance between two adjacent centres of the Gabor function, and the first wavelength value is lower than a second wavelength value and higher than or substantially equal to half the second wavelength value

wherein the models are characterized by a covariance matrix and by an average vector, and wherein the density of probability is calculated by the formula