Feature-based signatures for image identification
A method for deriving a representation of an image is described. The method involves processing signals corresponding to the image. At least one region of the image that is representative of the image is identified. In one embodiment, each identified region corresponds to a feature of the image. For each identified image region, a two dimensional function of at least part of the image region is derived and processed to obtain an intermediate representation of the at least part of the image region. A representation of the image is derived from the intermediate representation of at least one of the image regions.
1. A method of deriving a representation of an image by processing signals corresponding to the image, the method comprising:
identifying a plurality of regions of the image, wherein each region is representative of, or significant to, the image; and
for each identified image region:
deriving a function of at least part of the image region; and
using the function to obtain an intermediate representation of the at least part of the image region, and
deriving the representation of the image from the intermediate representations of two or more of the plurality of image regions, by
deriving a representation of each image region from the intermediate representation of the image region, and
using a plurality of derived representations of image regions to derive the representation of the image.
2. A method as claimed in claim 1 , comprising:
for each identified image region:
deriving a function of the image region, where the function of a translated, scaled or rotated version of the image is a translated or scaled version of the function of the image region.
3. A method as claimed in claim 1 , wherein the step of identifying a plurality of regions of the image comprises:
identifying features of the image, and
selecting image regions corresponding to identified features of the image.
4. A method as claimed in claim 3 , wherein the step of selecting image regions comprises:
selecting a plurality of said identified features of the image according to feature strength, and
determining image regions corresponding to the selected plurality of said identified features.
5. A method as claimed in claim 3 , wherein said step of selecting image regions comprises:
selecting a plurality of image regions corresponding to identified features of the image according to their position in the image.
6. A method as claimed in claim 5 , wherein the selected plurality of image regions are substantially distributed across the image.
7. A method as claimed in claim 5 , wherein each selected image region corresponds to a feature that is separated from the features of other selected image regions by at least a predetermined distance in the image.
8. A method as claimed in claim 3 , wherein said step of identifying features of the image uses an intensity image gradient method for detecting features.
9. A method as claimed in claim 1 , wherein:
the step of deriving a function of at least part of the image region comprises applying a first functional along lines over the image region;
the method further comprising:
applying a second functional to the derived function to obtain an intermediate representation of the at least part of the image region.
10. A method as claimed in claim 1 , wherein the intermediate representation is a one dimensional function.
11. A method as claimed in claim 1 , wherein the step of deriving a representation of each image region from the intermediate representation of the image region comprises:
processing the intermediate representation of the at least part of the image region using a plurality of frequency components of a frequency representation of the intermediate representation.
12. A method as claimed in claim 1 , wherein the step of deriving the representation of the image comprises:
combining selected parts of representations of a selected plurality of identified image regions derived from their corresponding intermediate representations.
13. A method as claimed in claim 1 , wherein the step of deriving the representation of the image from the intermediate representations of two or more of the plurality of image regions uses the intermediate representations of a predetermined number of image regions, wherein said predetermined number is preferably in the range of 2 to 256.
14. A method as claimed in claim 1 , further comprising:
associating geometric information relating to each image region with derived representations of the identified image regions.
15. A method for identifying an image, comprising:
deriving a representation of the image using the method of claim 1 , and
associating the representation with the image.
16. A method for comparing images comprising,
deriving a representation of each image by:
identifying a plurality of regions of the image, wherein each region is representative of, or significant to, the image;
for each identified image region:
deriving a function of at least part of the image region; and
using the function to obtain an intermediate representation of the at least part of the image region, and
deriving a representation of the image from the intermediate representations of two or more of the plurality of image regions, by deriving a representation of each image region from the intermediate representation of the image region, and using a plurality of derived representations of image regions to derive the representation of the image; and
comparing the derived representations.
17. A method for comparing images as claimed in claim 16 , comprising:
comparing image regions associated with the representation of a first image with image regions associated with the representation of a second image.
18. A method as claimed in claim 17 , comprising
using the representations of first and second images to determine a Hamming distance between a representation of an image region of the first image and a representation of an image region of the second image, and
identifying the first and second images as a potential match, if the Hamming distance between the representations of the corresponding image regions is below a first predetermined threshold.
19. A method as claimed in claim 18 , comprising determining a Hamming distance for representations of a plurality of image regions of the first and second images.
20. A method as claimed in claim 19 , further comprising:
identifying a predetermined number of pairs of image regions from the first and second images, the identified pairs having corresponding representations with a lowest determined Hamming distance which is below the first predetermined threshold.
21. A method as claimed in claim 20 , wherein each region in the identified pairs of image regions is unique.
22. A method as claimed in claim 19 , further comprising:
determining a cumulative Hamming distance using the sum of the Hamming distances for the representations of identified pairs of image regions, and
identifying the first and second images as a potential match, if the cumulative Hamming distance of the representations of identified pairs of image regions is below a second predetermined threshold.
23. A method as claimed in claim 18 , comprising:
applying a geometric constraint to geometric information for a plurality of pairs of image regions of pairs or images determined to be a possible match, to determine if the images are a match.
24. A method as claimed in claim 23 , wherein the step of applying a geometric constraint comprises:
determining a first ratio of distances between a predefined number of image regions in the first image;
determining a second ratio of distances between the predefined number of image regions in the second image; and
determining the first and second images as a match if a difference between the first and second ratios is below a third predetermined threshold.
25. A method as claimed in claim 24 , wherein the predefined number is three.
26. A method as claimed in claim 25 , wherein the step of applying a geometric constraint comprises:
for each of the first and second images:
calculating ratios of Euclidean distances between points (a, b, c) corresponding to the three image regions using the equations:
L
1
=
ab
ac
,
and
L
2
=
ab
bc
,
and
measuring a distance between the line length ratios (L i , L 2 ) of the first and second images (A, B) using the equations:
G
1
=
L
1
A
-
L
1
B
L
1
A
+
L
1
B
,
G
2
=
L
2
A
-
L
2
B
L
2
A
+
L
2
B
.
27. A method as claimed in claim 26 , wherein the third threshold includes a threshold for one or more of G 1 , G 2 , and G 1 +G 2 .
28. A method as claimed in claim 16 , comprising selecting matching images from the plurality of images based on a comparison of representations.
29. An apparatus for deriving a representation of an image by processing signals corresponding to the image or comparing images, the apparatus comprising:
at least one database storing a plurality images or representations of images: and
one or more processors adapted to execute the method of claim 1 or claim 16 .
30. Apparatus as claimed in claim 29 , further comprising:
a control device for controlling operation of the device one or more processors adapted to carry out the method of claim 1 or claim 16 .
31. Apparatus as claimed in claim 29 , further comprising:
a display means, and
an image selecting means.
32. A computer storage medium comprising instructions that, when executed by a computer, perform the method as claimed in claim 1 or claim 16 .