EP0704822A2
Expired / Lapsed
EP
⏱️ 0.0 years remaining
A method of storing and retrieving images of people, for example, in photographic archives and for the construction of identikit images
Assignee: Istituto Trentino di Cultura
•
Filing: 1995-09-01
•
Granted: —
•
Expires: 2015-09-01
Abstract
Abstract Each characteristic (or feature) of the images to be stored is associated with a region (r F ) the size of which corresponds to the average size of the feature in the images stored. If {F i } i = 1, ..., N is a set of images of the feature F, where F i is the image associated with each of the N people in the data base, on the basis of these images, a new set of images {Φ j } j=0,1, ..., N of the type F i = j=1 N C Fij Φ j + Φ o is generated, of which a subset {Φ} i=0, ..., K can be used to approximate, encode, compare and construct images of the type in question in an interactive manner. The preferred application is in techniques for searching photographic archives and for the construction of identikit images.
Inventors
- 👤 Roberto Brunelli
- 👤 Ornella Mich
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-11-09 14:09
CPC Classifications
G; G06; G06T; G06T11/00; G06F; G06F16/00; G06F16/50; G06F16/53; G06F16/532; G06V; G06V40/00; G06V40/10; G06V40/16
Claims Analysis
Total Claims: 42
Claim 1
Independent Claim
281 words
A method of storing a plurality (N) of images of
people, for example, in order to form photographic
archives or identikit systems in which the images to be
stored are distinguished by sets of features (F),
characterized in that it comprises the steps of: analysing the images, each of the features (F) having
a respective image (F i ) in each of the images, then
associating with each feature an average region (r F ) the
size of which corresponds to the average size of the
feature (F) in the plurality of images, generating from the respective images of (F i ) a set of
new images (Φ j ) such that: F i = j = 1 N C Fij Φ j + Φ 0 in which Φ o = 1 N i =1 N F i C Fij = ( Fi - o ) . j Φ j . Φ t = δ jt j , t = 1,..., N the new images Φ j in which 1 ≦ j ≦ N being selected as
eigenvectors of the covariance matrix in the set {F i - Φ₀}
so that a respective eigenvalue λ Fj can be associated with
each eigenvector, the set of new images (Φ j ) then being
arranged in a manner such that λ Fj ≧ λ Ft when j < t, storing the respective images in the form of vectors C Fi = { c Fil , ..., c Fik }, k < N so that a set of vectors, one for each of the regions of
the face used in the method, is associated with each of
the images of the plurality.
Claim 2
Dependent Claim
43 words
A method according to Claim 1, characterized in that,
in order to reconstruct each of the images from the
respective set of vectors, each image (F) is
reconstructed according to the equation F = j =1 K c Fj Φ j +Φ 0
Claim 3
Dependent Claim
36 words
A method according to Claim 2, characterised in that
the values of the co-ordinates of the vectors (C F ) are
selected by moving cursors to positions with which values
for the corresponding co-ordinates are associated.
Claim 4
Dependent Claim
73 words
A method according to Claim 2, characterized in that
the images (F) are reconstructed by operating on a region
(R F ) with larger dimensions than the average region (r F ),
the calculation of the covariance matrix and of the
expansion coefficients being limited to the average
region, the region outside the average region (r F ) and
inside the enlarged region (R F ) being used for connecting
the various features.
Claim 5
Independent Claim
45 words
A method according to any one of Claims 1 to 4,
characterized in that it comprises the step of reducing
the plurality of images by identifying amongst them the
images (F j ) the disparity (d) of which is less than a
predetermined threshold disparity.
Claim 6
Dependent Claim
67 words
A method according to Claim 5, characterized in that
the disparity (d) is calculated by an equation of the
type d 2 ( X , Y ) = h =1 T ( X h - Y h ) 2 where T represents the number of pixels of the image
considered as a vector, X and Y being the two images
between which the disparity (d) is calculated.
Claim 7
Dependent Claim
96 words
A method according to Claim 5, characterized in that
the disparity (d) is calculated according to the equation d 2 p = 1 - P xy in which P xy represents the standardized correlation
coefficient ρ XY = µ XY µ XX µ YY in which µ XY = 1 T h =1 T X h - 1 T j =1 T X j Y h - 1 T j =1 T Y j represents a second-order centred moment of the variables
X and Y representative of the images between which the
disparity (d) is calculated.
Claim 8
Independent Claim
55 words
A method according to any one of the preceding claims,
characterized in that the plurality of images (F) is
divided beforehand into subsets of images having
substantially similar features, for example, according to
data such as the race and sex of the people to whom the
images relate, the method being applied to these
separately.
Claim 9
Independent Claim
40 words
A method according to any one of the preceding claims,
characterized in that, before the analysis, the plurality
of images (F) is subjected to geometrical standardization
so that the images (F) of the plurality are of uniform
size and orientation.
Claim 10
Dependent Claim
35 words
A method according to Claim 9, characterized in that
the geometrical standardization operation comprises the
step of identifying the position of each pupil and the
operation of bringing the position to a predetermined
reference position.
Claim 11
Independent Claim
31 words
A method according to any one of the preceding
claims, characterized in that, before the analysis, the
images (F) of the plurality are subjected to a
standardization of their intensity values.
Claim 12
Dependent Claim
36 words
A method according to Claim 11, characterized in that
the standardization of the intensity values is carried
out by causing the average intensity value of a specific
region of the image to assume a predetermined value.
Claim 13
Independent Claim
68 words
A method according to any one of the preceding
claims, characterized in that, before the analysis, the
images (F) of the plurality are subjected to a
transformation of the type where M ' = I I * K G (σ) and the symbol * represents the convolution operation,
K G(σ) represents a Gaussian convolution kernel the
parameter σ (standard deviation) of which is correlated
with the interocular distance.
Claim 14
Dependent Claim
39 words
A method according to Claim 2, characterized in that,
at the reconstruction stage, the images (F) are
reconstructed as the sum of the features available and of
an average image corresponding to an arithmetic mean of
the images stored.
Claim 15
Independent Claim
152 words
A method according to any one of the preceding
claims, characterized in that it comprises the step of
associating, with each of the features, a respective
priority (P F ) and a respective map of values (W F ) the size
of which corresponds to that of the overall image, in
which are identified: a region (R F ) outside the feature, an inner region (r F ) of the feature, corresponding to
the region used for calculating the covariance matrix and
the coefficients, and a transition region (R F -r F ), as well as the step of setting the map values in the
region to 1 in the inner region (r F ), to 0 in the outer
region (R F ) and to an intermediate value (W F (x)) in the
transition region (R F -r F ), at the reproduction stage,
thus achieving a blending of the features.
Claim 16
Dependent Claim
53 words
A method according to Claim 15, characterized in that
the intermediate value W F (x) is selected in a one-dimensional
region by a law of the type: with the one-dimensional region defined as r F = [x 1 , x 2 ] R F = [x 0 , x 3 ] where x₀<x₁<x₂<x₃.
Claim 17
Dependent Claim
66 words
A method characterized in that, in the case of a two-dimensional
region, the product of two functions W F (x)
and W F (y) according to Claim 16 is used for one co-ordinate
and for the other co-ordinate, respectively, the
restored value being given by (...(((F 0 (1-W 1 )+W 1 F 1 )(1-W 2 )+W 2 F 2 )(1-W 3 )+W 3 F 3 )...)
Claim 18
Independent Claim
89 words
A method according to any one of the preceding
claims, comprising the step of comparing the images
stored with an additional image (X) in the course of
formation, generating a respective similarity parameter,
then sorting the set of images stored in order of
increasing disparity or dissimilarity values (D) with
respect to the image (T i ) in the course of construction,
as well as the step of displaying a certain number of
stored images having a certain degree of similarity to
the images in the course of construction.
Claim 19
Dependent Claim
129 words
A method according to Claim 18, characterized in that
the disparity or dissimilarity (D) is defined according
to an equation of the type D ( X , I i ) = F =1 Q a F ω F | C F - C Fi | 2 where Q is the number of features which can be selected;
a F = 1 or 0 according to whether the feature has been
selected or not and the quantity ω F : ω F = i =1 K λ 2 Fi -1 represents a standardization factor with which to weight
the various features and the vectors C F , C Fi represent the
image being constructed and the image i of the data base,
respectively, within the limits of the feature F.
Claim 20
Dependent Claim
36 words
A method according to Claim 19, characterized in that
pointing to (selecting) one of the images displayed
replaces the descriptive components of the features
selected in the image being constructed with those of the
image selected.
Claim 21
Dependent Claim
55 words
A method according to Claim 19, characterized in that
the set of images in the data base is organized and
presented in a hierarchical manner, and in that pointing
to (selecting) one of the images displayed replaces the
descriptive components of the features selected in the
image being constructed with those of the image selected.
Claim 22
Independent Claim
286 words
A system for storing a plurality (N) of images of
people, for example, in order to form photographic
archives or identikit systems, in which the images to be
stored are distinguished by sets of features (F),
characterized in that it comprises: acquisition and analysis means (C, 5; 2) for analysing
the images, each of the features (F) having a respective
image (F i ) in each of the images, then associating with
each feature an average region (r F ), the size of which
corresponds to the average size of the feature (F) in the
plurality of images, processor means (4) for generating, from the respective
images (F i ) a set of new images (Φ j ) such that F i = j = 1 N C Fij Φ j + Φ 0 where Φ o = 1 N i =1 N F i C Fij = ( F i - Φ o ) Φ j Φ j . Φ t = δ jt j , t = 1,..., N the new images Φ j in which 1 ≦ j ≦ N being selected as
eigenvectors of the covariance matrix in the set {F i - Φ₀}
so that a respective eigenvalue λ Fj can be associated with
each eigenvector, the processing means (4) then sorting
the set of new images (Φ j ) in a manner such that λ Fj ≧ λ Ft when j < t and storing the respective images in the form
of vectors C Fi = {C Fil , ..., C Fik } so that a set of vectors, one for each of the regions of
the face used in the system, is associated with each
image of the plurality.
Claim 23
Dependent Claim
58 words
A system according to Claim 22, characterized in that
it comprises restoring means (4, 5) for reconstructing
each of the images from the respective set of vectors,
the restoring means (4, 5) reconstructing each image (F)
according to the equation F i = j = 1 k C Fij Φ j + Φ 0 k < N ;
Claim 24
Dependent Claim
39 words
A system according to Claim 23, characterized in that
it comprises cursor means for selecting the co-ordinates
of the vectors ( C Fi ), the position of the cursor means
being associated with a value for the corresponding
coordinate.
Claim 25
Dependent Claim
70 words
A system according to Claim 23, characterized in that
the restoring means operate on a region (R F ) with larger
dimensions than the average region (r F ), the calculation
of the covariance matrix and of the expansion
coefficients being limited to the average region, the
region outside the average region (r F ) and inside the
enlarged region (R F ) being used for connecting the
various features.
Claim 26
Independent Claim
41 words
A system according to any one of Claims 22 to 25,
characterized in that the processing means (4) reduce the
plurality of images by identifying amongst them the
images the disparity (d) of which is less than a
predetermined threshold disparity.
Claim 27
Dependent Claim
70 words
A system according to Claim 26, characterized in that
the processing means (4) calculate the disparity (d) with
an equation of the type d 2 ( X , Y ) = h =1 T ( X h - Y h ) 2 where T represents the number of pixels of the image
considered as a vector, X and Y being the two images
between which the disparity (d) is calculated.
Claim 28
Dependent Claim
100 words
A system according to Claim 26, characterized in that
the processing means (4) calculate the disparity (d)
according to the equation d 2 p = 1 - P xy where P xy represents the standardized correlation
coefficient ρ XY = µ XY µ XX µ YY in which µ XY = 1 T h =1 T X h - 1 T j =1 T X j Y h > - 1 T j =1 T Y j represents a second order centred moment of the variables
X and Y representative of the images between which the
disparity (d) is calculated.
Claim 29
Independent Claim
58 words
A system according to any one of Claims 22 to 28,
characterized in that the processing means (4) divide the
plurality of images (F) into subsets of images having
substantially similar features, for example, according to
data such as the race and sex of the people to whom the
images relate, the method being applied to these
separately.
Claim 30
Independent Claim
61 words
A system according to any one of the preceding claims
22 to 29, characterized in that it comprises pre-processing
means (3) which act between the acquisition
and analysis means (2) and the processing means (4), and
in which the plurality of images is subjected to
geometrical standardization so that the images (F) of the
plurality are of uniform size and orientation.
Claim 31
Dependent Claim
35 words
A system according to Claim 30, characterized in that
the pre-processing means (3) identify the position of
each pupil in the images (F), the pre-processing means
(3) bringing this position to a predetermined reference
position.
Claim 32
Independent Claim
53 words
A system according to any one of the preceding claims
22 to 31, characterized in that it comprises pre-processing
means (3) which act between the acquisition
and analysis means (2) and the processing means (4), and
in which the images (F) of the plurality are subjected to
a standardization of their intensity values.
Claim 33
Dependent Claim
37 words
A system according to Claim 32, characterized in
that, during the standardization of the intensity values,
the pre-processing means (3) cause the average intensity
value of a specific region of the image to assume a
predetermined value.
Claim 34
Independent Claim
90 words
A system according to any one of the preceding Claims
22 to 33, characterized in that it comprises pre-processing
means (3) which act between the acquisition
and analysis means (2) and the processing means (4), and
in which the images (F) of the plurality are subjected to
a transformation of the type where M ' = I I * K G (σ) and the symbol * represents the convolution operation,
K G(σ) represents a Gaussian convolution kernel the
parameter σ (standard deviation) of which is correlated
to the interocular distance.
Claim 35
Dependent Claim
38 words
A system according to Claim 23, characterized in that
the restoring means (4, 5) reconstruct the images as the
sum of the features available and of an average image
corresponding to an arithmetic mean of the images stored.
Claim 36
Independent Claim
162 words
A system according to any one of the preceding claims
22 to 35, characterized in that the processing means (4)
associate with each of the features a respective priority
(P F ) and a respective map of values (W F ) the size of which
corresponds to that of the overall image, in which are
identified: a region (R F ) outside the feature, an inner region (r F ) of the feature (RA F ) corresponding
to the region used for calculating the covariance matrix
and the coefficients, and a transition region (R F - r F ), and in that the processing means (4) set the map values
in the region to 1 in the inner region (r F ), to 0 in the
outer region (R F ) and to an intermediate value (W F (x)) in
the transition region (R F - r F ) so that the restoring
means (4, 5) achieve a blending of the features.
Claim 37
Dependent Claim
46 words
A system according to Claim 36, characterized in that
the intermediate value is selected by a law of the type: with the one-dimensional region defined as r F = [x 1 , x 2 ] R F = [x 0 , x 3 ] where x₀<x₁<x₂<x₃.
Claim 38
Dependent Claim
72 words
A system according to Claim 37, characterized in
that, in the case of a two-dimensional region, the
processing means use the product of two functions W F (x)
and W F (y) according to Claim 37 for one co-ordinate and
for the other co-ordinate, respectively, the restored
value being given by (...(((F 0 (1-W 1 )+W 1 F 1 )(1-W 2 )+W 2 F 2 )(1-W 3 )+W 3 F 3 )...)
Claim 39
Dependent Claim
89 words
A system according to Claim 23, characterized in that
the processing means (4) compare the images stored with
an additional image (X) in course of formation,
generating a respective similarity parameter then sorting
the set of images stored in order of increasing values of
disparity or dissimilarity (D) with respect to the images
(I i ) in course of construction, the system also comprising
restoring means (4, 5) for displaying a certain number of
stored images having a certain degree of similarity to
the images in course of construction.
Claim 40
Dependent Claim
129 words
A system according to Claim 39, characterized in that
the disparity or dissimilarity (D) is defined according
to an equation of the type D ( X , I i ) = F =1 Q a F ω F | C F - C Fi | 2 where Q is the number of features which can be selected,
a F = 1 or 0 according to whether the feature has been
selected or not, and the quantity ω F : ω F = i =1 K λ 2 Fi -1 represents a standardization factor with which to weight
the various features and the vectors C F , C Fi represent the
image being constructed and the image i of the data base,
respectively, within the limits of the feature F.
Claim 41
Dependent Claim
36 words
A system according to Claim 40, characterized in that
pointing to (selecting) one of the images displayed
replaces the descriptive components of the features
selected in the image being formed with those of the
image selected.
Claim 42
Dependent Claim
55 words
A system according to Claim 40, characterized in that
the set of images of the data base is organized and
presented in a hierarchical manner, and in that pointing
to (selecting) one of the images displayed replaces the
descriptive components of the features selected in the
image being constructed with those of the image selected.
References Cited (Backward Citations)
No citation records indexed.
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.