Face recognition system and method using face pattern words and face pattern bytes

• US 20120257800A1
• Filed: 04/05/2012
• Published: 10/11/2012
• Est. Priority Date: 04/05/2011
• Status: Active Grant

• Current Assignee: Alcorn State University
• Original Assignee: Alcorn State University
• Inventors: Zheng, Yufeng
• Granted Patent: US 8,917,914 B2
• Time in Patent Office: Days
• US Class Current: 382/118
• CPC Class Codes: G06V 10/52 Scale-space analysis, e.g. wavelet analysis multi-scale boundary representations G06V10/42
• CPC Class Codes: G06V 40/169 Holistic features and representations, i.e. based on the facial image taken as a whole

Abstract

The present invention provides a novel system and method for identifying individuals and for face recognition utilizing facial features for face identification. The system and method of the invention comprise creating facial features or face patterns called face pattern words and face pattern bytes for face identification. The invention also provides for pattern recognitions for identification other than face recognition. The invention further provides a means for identifying individuals based on visible and/or thermal images of those individuals by utilizing computer software implemented by instructions on a computer or computer system and a computer readable medium containing instructions on a computer system for face recognition and identification.

Claims

1. A method for face recognition and identification, the method comprising:

receiving at least one probe face image and at least one gallery face image of a subject to be identified, wherein the at least one probe face image is a real-time image and the at least one gallery face image is an image from a pre-existing image database and not a real-time image;

preprocessing the at least one probe face image and the at least one gallery face image;

normalizing the at least one probe face image and the at least one gallery face image to obtain a consistent dynamic image range;

creating at least one set of face patterns comprising face pattern words (FPWs) or face pattern bytes (FPBs) using at least one set of Gabor wavelet transform (GWT) coefficients with each probe face image and gallery face image;

calculating the Hamming Distance between the at least one probe face image and the at least one gallery face image using the FPWs or FPBs and a face mask; and

identifying a subject by using the FPWs or FPBs,

wherein the method utilizes at least one processor for receiving, preprocessing, and normalizing the images, for creating the at least one set of face patterns comprising the FPWs or FPBs, for calculating the Hamming Distance, and for identifying a subject.

2. The method of claim 1, wherein the at least one probe face image is a thermal face image or a visible face image and wherein the at least one gallery face image is a thermal face image or a visible face image.

3. The method of claim 1, wherein the subject is identified using the smallest Hamming Distance between the at least one probe face image and the at least one gallery face image.

4. The method of claim 3, wherein the at least one probe face image and the at least one gallery face image are both visible face images or the at least one probe face image and the at least one gallery face image are both thermal face images, and wherein the at least one gallery face image is of one or more subjects and is stored in a database prior to face recognition.

5. The method of claim 3, wherein the FPWs or FPBs are masked with the face mask in the calculation of the Hamming Distance to minimize or eliminate image distractions.

6. The method of claim 1, wherein the creation of FPWs is pixel by pixel for each probe and gallery face image and comprises using at least one configuration of the GWT coefficients.

7. The method of claim 6, wherein the creation of FPWs comprises the maximal magnitude of the GWT coefficients of sixteen orientations at each frequency band of a total of eight frequency bands.

8. The method of claim 7, wherein the eight central frequencies of the eight frequency bands are 0.0808, 0.0948, 0.1113, 0.1306, 0.1532, 0.1798, 0.2110, and 0.2475, and wherein the frequencies are normalized spatial frequencies.

9. The method of claim 7, wherein the creation of FPWs comprises eight sets of 4-bit orientation bit code with optimized binary coding and wherein each set of 4-bit orientation bit code corresponds to the index of a selected orientation at each frequency band.

10. The method of claim 9, wherein the orientation bit code with optimized binary coding can be used for other pattern recognitions.

11. The method of claim 1, wherein the creation of FPBs is pixel by pixel for each probe and gallery face image and comprises using at least one configuration of the GWT coefficients.

12. The method of claim 1, wherein the creation of FPBs is pixel by pixel for each probe and gallery face image and comprises using at least one FPW.

13. The method of claim 1, wherein the creation of FPBs comprises the maximal magnitude of the GWT coefficients of sixteen orientations at each frequency band of a total of eight frequency bands.

14. The method of claim 13, wherein the eight central frequencies of the eight frequency bands are 0.0808, 0.0948, 0.1113, 0.1306, 0.1532, 0.1798, 0.2110, and 0.2475, and wherein the frequencies are normalized spatial frequencies.

15. The method of claim 13, wherein the creation of FPBs comprises two sets of 4-bit orientation bit code with optimized binary coding.

16. The method of claim 15, wherein each set of 4-bit orientation bit code corresponds to the index of a selected orientation at each frequency band and wherein the selected orientation is the mode of low four bands or high four bands.

17. A method for face recognition and identification, the method comprising:

receiving at least one probe face image and at least one gallery face image of a subject to be identified, wherein the at least one probe face image is a real-time image and the at least one gallery face image is an image from a pre-existing image database and not a real-time image;

preprocessing the at least one probe face image and the at least one gallery face image;

normalizing the at least one probe face image and the at least one gallery face image to obtain a consistent dynamic image range;

creating at least one set of face patterns comprising face pattern words (FPWs) or face pattern bytes (FPBs) using at least one set of Gabor wavelet transform (GWT) coefficients with each probe face image and gallery face image,

wherein the at least one probe face image is a thermal face image or a visible face image and wherein the at least one gallery face image is a thermal face image or a visible face image, and

wherein the creation of FPWs and FPBs is pixel by pixel for each probe and gallery face image and comprises using at least one configuration of the GWT coefficients and further comprises eight sets and two sets, respectively, of 4-bit orientation bit code with optimized binary coding;

calculating the Hamming Distance between the at least one probe face image and the at least one gallery face image using the FPWs or FPBs and a face mask;

optimizing the orientation bit code by minimizing the Hamming Distance between all neighboring orientations and maximizing the Hamming Distance between all orthogonal orientations; and

identifying a subject by using the FPWs or FPBs,

wherein the method utilizes at least one processor for receiving, preprocessing, and normalizing the images, for creating the at least one set of face patterns comprising the FPWs or FPBs, for calculating the Hamming Distance, for optimizing the orientation bit code, and for identifying a subject.

18. The method of claim 17, wherein the number of neighboring orientations is two or more.

20. A computer-readable medium comprised of computer processor-based instructions that utilize the method of claim 17 to instruct a computer-based device for face recognition and identification.

21. The medium of claim 20, wherein the at least one probe face image is a thermal face image or a visible face image and wherein the at least one gallery face image is a thermal face image or a visible face image.

22. The medium of claim 21, wherein the subject is identified using the smallest Hamming Distance between the at least one probe face image and the at least one gallery face image.

23. The medium of claim 22, wherein the at least one probe face image and the at least one gallery face image are both visible face images or the at least one probe face image and the at least one gallery face image are both thermal face images, and wherein the at least one gallery face image is of one or more subjects and is stored in a database prior to face recognition.

24. The medium of claim 22, wherein the FPWs or FPBs are masked with the face mask in the calculation of the Hamming Distance to minimize or eliminate image distractions.

25. The medium of claim 20, wherein the creation of FPWs is pixel by pixel for each probe and gallery face image and comprises using at least one configuration of the GWT coefficients.

26. The medium of claim 25, wherein the creation of FPWs comprises the maximal magnitude of the GWT coefficients of sixteen orientations at each frequency band of a total of eight frequency bands.

27. The medium of claim 26, wherein the eight central frequencies of the eight frequency bands are 0.0808, 0.0948, 0.1113, 0.1306, 0.1532, 0.1798, 0.2110, and 0.2475, and wherein the frequencies are normalized spatial frequencies.

28. The medium of claim 26, wherein the creation of FPWs comprises eight sets of 4-bit orientation bit code with optimized binary coding and wherein each set of 4-bit orientation bit code corresponds to the index of a selected orientation at each frequency band.

29. The medium of claim 28, wherein the orientation bit code with optimized binary coding can be used for other pattern recognitions.

30. The medium of claim 20, wherein the creation of FPBs is pixel by pixel for each probe and gallery face image and comprises using at least one configuration of the GWT coefficients.

31. The medium of claim 20, wherein the creation of FPBs is pixel by pixel for each probe and gallery face image and comprises using at least one FPW.

32. The medium of claim 20, wherein the creation of FPBs comprises the maximal magnitude of the GWT coefficients of sixteen orientations at each frequency band of a total of eight frequency bands.

33. The medium of claim 32, wherein the eight central frequencies of the eight frequency bands are 0.0808, 0.0948, 0.1113, 0.1306, 0.1532, 0.1798, 0.2110, and 0.2475, and wherein the frequencies are normalized spatial frequencies.

34. The medium of claim 32, wherein the creation of FPBs comprises two sets of 4-bit orientation bit code with optimized binary coding.

35. The medium of claim 34, wherein each set of 4-bit orientation bit code corresponds to the index of a selected orientation at each frequency band and wherein the selected orientation is the mode of low four bands or high four bands.

36. The medium of claim 20, wherein the number of neighboring orientations is two or more.

19. A system for face recognition and identification, the system comprising:

means for receiving at least one probe face image and at least one gallery face image of a subject to be identified, wherein the at least one probe face image is a real-time image and the at least one gallery face image is an image from a pre-existing image database and not a real-time image;

a pre-existing image database containing the at least one gallery face image;

means for preprocessing the at least one probe face image and the at least one gallery face image;

means for normalizing the at least one probe face image and the at least one gallery face image to obtain a consistent dynamic image range;

means for creating at least one set of face patterns comprising face pattern words (FPWs) or face pattern bytes (FPBs) using at least one set of Gabor wavelet transform (GWT) coefficients with each probe face image and gallery face image,

wherein the at least one probe face image is a thermal face image or a visible face image and wherein the at least one gallery face image is a thermal face image or a visible face image, and

wherein the creation of FPWs and FPBs is pixel by pixel for each probe and gallery face image and comprises using at least one configuration of the GWT coefficients and further comprises eight sets and two sets, respectively, of 4-bit orientation bit code with optimized binary coding;

means for calculating the Hamming Distance between the at least one probe face and the at least one gallery face using the FPWs or FPBs and a face mask;

means for optimizing the orientation bit code by minimizing the Hamming Distance between all neighboring orientations and maximizing the Hamming Distance between all orthogonal orientations; and

means for identifying a subject by using the FPWs or FPBs,

wherein the system utilizes at least one processor for the means for receiving, preprocessing, and normalizing the images, for the means for creating the at least one set of face patterns comprising the FPWs or FPBs, for the means for calculating the Hamming Distance, for the means for optimizing the orientation bit code, and for the means for identifying a subject.