Object class identification, verification of object image synthesis

1. A method of identifying an object class using a model based upon appearance parameters derived by comparing images of objects of different classes, the model including a representation of a probability density function describing a range over which appearance parameters may vary for a given class of object, the model further including a defined relationship between the shape of the probability density function and the location of the probability density function in appearance space, wherein the method comprises:

generating appearance parameters representative of an unknown object,

estimating an appropriate probability density function for the unknown object based upon the location of the appearance parameters in appearance space and using the defined relationship between the shape of the probability density function and the location of the probability density function in appearance space,

then iteratively modifying at least some of the appearance parameters within limits determined using the probability density function to provide a set of appearance parameters which identify an object class.

2. A method according to claim 1 , wherein a threshold level of the probability density function is determined, and the appearance parameters are constrained to have a probability density greater than the threshold.

3. A method according to claim 1 , wherein the objects are faces, and a given class of object is a face having a particular identity.

4. A method according to claim 1 , wherein the relationship between the appearance parameters and the probability density function is a relationship between average appearance parameters determined for each class of object, and probability density functions associated with each class of object.

5. A method according to claim 1 , wherein the probability density function is a Gaussian function.

6. A method according to claim 1 , wherein the probability density function is approximated as a Gaussian with a given covariance matrix.

7. A method according to claim 1 , wherein the model is the Active Appearance Model.

8. A method of verifying the identity of an object class using a model based upon appearance parameters derived by comparing images of objects of different classes, the model including a representation of a probability density function describing the variation of appearance parameters for a given class of object, the model function including a defined relationship between the shape of the probability density function and the location of the probability density function in appearance space, wherein for an object class for which verification will subsequently be required, the method comprises:

using a series of images of that object class to generate a representation of a specific probability density function describing the variation of appearance parameters for that object class, and

during subsequent verification of the object class, comparing an image of an object of an unknown class with the object of known class by generating appearance parameters representative of the object of unknown class and iteratively modifying at least some of the appearance parameters within limits determined using the specific probability density function, to provide a set of appearance parameters which are compared with a predetermined threshold to provide verification.

9. A method according to claim 8 , wherein a threshold level of the probability density function is determined, and the appearance parameters are constrained to have a probability density greater than the threshold.

10. A method according to claim 8 , wherein the objects are faces, and a given class of object is a face having a particular identity.

11. A method according to claim 8 , wherein the relationship between the appearance parameters and the probability density function is a relationship between average appearance parameters determined for each class of object, and probability density functions associated with each class of object.

12. A method according to claim 8 , wherein the probability density function is a Gaussian function.

13. A method according to claim 8 , wherein the probability density function is approximated as a Gaussian with a given covariance matrix.

14. A method according to claim 8 , wherein the model is the Active Appearance Model.

15. A method of generating a synthesized image of an object class using a model based upon appearance parameters derived by comparing images of objects of different classes, the model including a representation of a probability density function describing the variation of appearance parameters for a given class of object, the model further including a defined relationship between the shape of the probability density function and the location of the probability density function in appearance space, wherein for an object class to be synthesized, the method comprises:

using a series of images of that object class to generate a representation of a specific probability density function describing the variation of appearance parameters for that object class, and

during subsequent synthesis of the object class, confining the appearance parameters used within limits determined using the specific probability density function.

16. A method according to claim 15 , wherein a threshold level of the probability density function is determined, and the appearance parameters are constrained to have a probability density greater than the threshold.

17. A method according to claim 15 , wherein the objects are faces, and a given class of object is a face having a particular identity.

18. A method according to claim 15 , wherein the relationship between the appearance parameters and the probability density function is a relationship between average appearance parameters determined for each class of object, and probability density functions associated with each class of object.

19. A method according to claim 15 , wherein the probability density function is a Gaussian function.

20. A method according to claim 15 , wherein the probability density function is approximated as a Gaussian with a given covariance matrix.

21. A method according to claim 15 , wherein the model is the Active Appearance Model.