Method and a system for organizing an image database

1. Method for organizing a set of images into subsets of images, the method comprising the procedures of:

for each of said images, producing a respective model according to a plurality of feature points and the geometric relations between said feature points;

determining a similarity index between each pair of said images, according to said respective model of each image of said pair of said images;

producing a distance matrix, according to said similarity index between each pair of said images and determining a set of eigenvectors of said distance matrix;

associating each of said images with respective image coordinates in a two-dimensional Euclidean space thereby embedding said set of images within said two-dimensional Euclidean space, said respective image coordinates of each of said images being determined according to said set of eigenvectors of said distance matrix, said image coordinates of said images representing the inner structure of said images; and

sorting said images, according to said image coordinates of each of said images.

2. The method according to claim 1 , further comprising a preliminary procedure of digitizing said set of images and storing said set of images on a database.

3. The method according to claim 1 , wherein said procedure of determining said similarity index, includes the sub-procedures of:

forming an affinity matrix A according to distances between pairs of feature points in said pair of images;

forming a set of eigenvectors of said affinity matrix A, and reshaping said eigenvectors of said affinity matrix A to form a matrix of eigenvectors of predetermined dimensions N 1 ×N 2 ;

creating a new empty matrix Z;

determining which entry of said matrix of eigenvectors is of the largest magnitude, and denoting the coordinates of said largest magnitude entry;

setting to the value of one, an entry of said new matrix Z, having the same coordinates as those of said largest magnitude entry;

erasing from said matrix of eigenvectors the row and the column, corresponding to the coordinates of said largest magnitude entry;

repeating said procedures of determining largest magnitude entry, setting to one, and erasing the row and the column, until all of the entries of said matrix of eigenvectors are erased; and

determining a similarity index according to the following equation:

Similarity

=

Z

T

⁢

AZ

N

1

.

4. The method according to claim 3 , wherein N1 is a number of interest points in a first image of said pair of images and N2 is a number of interest points in a second image of said pair of images.

5. The method according to claim 3 , wherein said procedure of producing a distance matrix between each pair of said images includes the sub-procedures of:

determining a similarity matrix S according to a set of similarity indexes between each pair of said set of images;

determining a distance matrix D according to said similarity matrix S:

D= 1./ S

D ( i,i )=0;

determining a general affinity matrix G according to said distance matrix D:

G =exp[− D /sigma];

determining a matrix of eigenvectors U of said general affinity matrix G;

denoting the entries of the first two columns of said eigenvectors matrix U as the coordinates of said images; and

identifying at least one said subset of images according to said coordinates of said images.

6. System for organizing a set of images into subsets of images, according to similarity of the content of the images, and for organizing series of panoramic images in a correct order thereof, the system comprising:

a database storing said set of images; and

a processor, wherein said processor produces a respective model, for each of said images according to a plurality of feature points and the geometric relations between said feature points, said processor determines a similarity index between each pair of said images according to said respective model of each image of said pair of said images, said processor produces a distance matrix according to said similarity index between each pair of said images and determines a set of eigenvectors of said distance matrix, said processor associates each of said images with respective image in a two-dimensional Euclidean space thereby embedding said set of images within said two-dimensional Euclidean space, said processor determines said respective image coordinates of each of said images according to said set of eigenvectors of said distance matrix, said image coordinates of said images represent the inner structure of said images, said processor sorts said images according to said respective image coordinates of each of said images.

7. The system according to claim 6 , further comprising an image digitizer coupled with said database for digitizing said set of images and storing said set of images on said database.