Visual object appearance modelling using image processing

1. A computer-implemented method of generating a model from a set of images, the method comprising:

processing a plurality of data items, each data item representing an image of said set of images, to determine variability between said plurality of data items;

generating model data representing said model based upon said data items and said variability, wherein the influence of each of said data items upon the generated model is determined by a relationship between a respective one of said data items and said variability.

2. A method according to claim 1 , wherein each of said plurality of data items is a normalized data item.

3. A method according to claim 2 , further comprising:

receiving a plurality of initial data items, each initial data item comprising a plurality of elements;

determining a median value for each element based upon values of that element in each of said plurality of initial data items.

4. A method according to claim 3 , further comprising:

determining a median absolute deviation for each element based upon values of that element in each of said plurality of initial data items.

5. A method according to claim 4 , further comprising:

processing each element of a respective initial data item with reference to a respective median value and median absolute deviation value to determine a normalization factor for the respective initial data item; and

generating a respective normalized data item by applying said normalization factor to the respective initial data item.

6. A method according to claim 1 , wherein generating said model based upon said data items comprises associating a weight with each of said data items, said weights being based upon said variability.

7. A method according to claim 6 , further comprising computing a weight for each of said data items.

8. A method according to claim 7 , wherein computing a weight for a respective data item comprises determining a relationship between the respective data item and an average data item.

9. A method according to claim 8 , wherein said average data item is a median data item.

10. A method according to claim 8 , further comprising determining a value indicating a confidence in each of said data items.

11. A method according to claim 10 , wherein the confidence of a respective data item is based upon the relationship between the respective data item and the average data item and variability of said data items.

12. A method according to claim 10 , further comprising generating a probability distribution from said values indicating a confidence in each of said data items, and determining a probability value for each of said data items based upon the generated distribution.

13. A method according to claim 12 , wherein the weight associated with a particular data item is a function of the determined probability value.

14. A method according to claim 13 , wherein said function provides a first weight value when said probability is within a first range, and said function returns a weight value in a second range when said probability is in a second range.

15. A method according to claim 1 , wherein said model is a statistical model.

16. A method according to claim 1 , wherein each data item is a vector.

17. A method according to claim 1 , wherein each element of each data item represents a pixel value.

18. A non-transitory computer readable storage medium carrying a computer program comprising computer readable instructions configured to cause a computer to carry out a method according to claim 1 .

19. A non-transitory computer readable storage medium storing data representing a model created using the method of claim 1 .

20. A computer apparatus for generating a model from a set of images, the computer apparatus comprising:

a memory storing processor readable instructions; and

a processor arranged to read and execute instructions stored in said memory;

wherein said processor readable instructions comprise instructions arranged to control the computer to carry out a method according to claim 1 .

21. A computer-implemented method of generating a model from a set of images, the method comprising:

processing a plurality of data items, each data item representing an image of said set of images, to determine variability between said plurality of data items;

generating model data representing said model based upon said data items and said variability, wherein the influence of each of said data items upon the generated model is determined by a relationship between a respective one of said data items and said variability;

wherein generating said model comprises:

associating a weight with each of said data items, said weights being based upon said variability;

determining a difference relationship between each of said data items and an average data item to generate a plurality of difference data items;

applying said weights to said difference data items; and

generating a matrix indicating how each weighted difference data item varies with reference to each other weighted difference data item.

22. A method according to claim 21 , wherein generating said model further comprises:

performing eigensystem analysis on said matrix to generate a plurality of eigenvectors; and

creating a matrix containing at least some of said plurality of eigenvectors.

23. A method according to claim 22 , wherein generating said plurality of eigenvectors further comprises generating a plurality of eigenvalues, wherein said plurality of eigenvalues are generated by processing a further matrix, said further matrix being of smaller size than said matrix.

24. A computer-implemented method of generating a model from a set of images, the method comprising:

processing a plurality of data items, each data item representing an image of said set of images, to determine variability between said plurality of data items;

generating model data representing said model based upon said data items and said variability, wherein the influence of each of said data items upon the generated model is determined by a relationship between a respective one of said data items and said variability, wherein said model is a statistical model] having a form:

m+Pb

where m is an average data item;

P is a matrix indicating permissible variation; and

b is an example instance of the model.

25. A computer implemented method of generating data indicating how well an input image satisfies a model representing a set of images, the method comprising:

processing an input data item representing the input image to generate data indicating how well the image can be represented by the model;

wherein the input data item comprises a plurality of elements, each element representing a respective image element of the input image, and processing the input data item to determine how well the input image satisfies the model comprises:

determining an influence which each element of the input data item should have upon the generated data, and generating said data such that each element of the input data item has the determined influence.

26. A method according to claim 25 , wherein determining an influence which an element of the input data item should have upon the generated data comprises:

generating a weight data item for each element of the input data item.

27. A method according to claim 26 , wherein each of said weight data items is generated using iterative processing.

28. A method according to claim 25 further comprising generating output data indicating the influence of at least some of said input elements upon said determination, wherein said output data comprises graphical output data indicating pixels of the input image which contribute relatively little to the determination.

29. A method according to claim 25 , wherein said data indicating how well an input image satisfies a model is generated iteratively.

30. A method according to claim 25 , wherein the input data item is a normalized input data item.

31. A non-transitory computer readable storage medium carrying a computer program comprising computer readable instructions configured to cause a computer to carry out a method according to claim 25 .

32. A computer implemented method of generating data indicating how well an input image satisfies a model representing a set of images, the method comprising:

processing an input data item representing the input image to generate data indicating how well the image can be represented by the model;

wherein the input data item comprises a plurality of elements, each element representing a respective image element of the input image, and processing the input data item to determine how well the input image satisfies the model comprises:

determining an influence which each element of the input data item should have upon the generated data, and generating said data such that each element of the input data item has the determined influence,

wherein generating said data such that each element has the determined influence comprises:

determining a relationship between each element in a plurality of sample data items representing images of the set of images and a respective element of the input data item; and

applying a respective one of said weight data items to each relationship to generate said data indicating how well the input image satisfies the model.

33. A computer implemented method of generating data indicating how well an input image satisfies a model representing a set of images, the method comprising:

processing an input data item representing the input image to generate data indicating how well the image can be represented by the model;

wherein the input data item comprises a plurality of elements, each element representing a respective image element of the input image, and processing the input data item to determine how well the input image satisfies the model comprises:

determining an influence which each element of the input data item should have upon the generated data, and generating said data such that each element of the input data item has the determined influence, wherein determining an influence which an element of the input data item should have upon the generated data comprises:

generating a weight data item for each element of the input data item,

wherein generating each weight data item comprises determining a relationship between an element of the input data item and the variability of a respective element of a plurality of sample data items, wherein the relationship between an element of the input data item and the variability of an element of a plurality of sample data items is a relationship between the difference between the element of the input data item and an average element value, and said variability.