1. An object verification method, comprising:
obtaining a target Chi square inner product based on a target local binary pattern (LBP) feature of a target image and a target kernel LBP feature to generate a target image feature vector;
projecting the target image feature vector to a kernel fisher discriminant analysis (KFDA) basis vector;
obtaining a query Chi square inner product based on a query LBP feature of a query image and a query kernel LBP feature to generate a query image feature vector;
projecting the query image feature vector to the KFDA basis vector; and
providing a similarity degree between the target image and the query image based on the projected target image feature vector and the projected query image feature vector and outputting the similarity degree as a result of verification between the target image and the query image for object verification.
2. The method of claim 1, wherein the target kernel LBP feature and query kernel LBP feature are a same kernel LBP feature.
3. The method of claim 1, further comprising generating the KFDA basis vector by using at least one training LBP feature of at least one training image.
4. The method of claim 3, wherein the generating of the KFDA basis vector comprises:
obtaining the at least one training LBP feature of the at least one image;
obtaining the training kernel LBP feature;
obtaining a training Chi square inner product by using the at least one training LBP feature and the training kernel LBP feature; and
obtaining the KFDA basis vector by performing a KFDA for the training Chi square inner product.
5. The method of claim 4, wherein the training kernel LBP feature, the target kernel LBP feature, and query kernel LBP feature are a same kernel LBP feature.
6. The method of claim 4, wherein in the obtaining of the similarity degree, the similarity degree is determined by comparing Euclidian distances.
7. The method of claim 4, wherein at least one of the training Chi square inner product, the target Chi square inner product, and the query Chi square inner product is obtained by performing a following kernel equation:
\u2062
k
\u2061
(
x
,
y
)
=
exp
(
–
\u03c7
2
\u2061
(
x
,
y
)
2
\u2062
\u03c3
2
)
\u2062
\u2062
where
\u03c7
2
\u2061
(
x
,
y
)
\u2062
\u2062
is
\u2062
\u2062
defined
\u2062
\u2062
by
\u2062
\u2062
\u03c7
2
\u2061
(
S
,
M
)
=
\u2211
i
\u2062
(
S
i
–
M
j
\u2062
)
2
S
i
+
M
i
.
8. The method of claim 4, wherein in the obtaining of the at least one training LBP feature of the at least one training image, the at least one training LBP feature is expressed as a corresponding LBP histogram by performing a corresponding LBP operation for the at least one training image.
9. The method of claim 1, wherein at least one of the target Chi square inner product and the query Chi square inner product is obtained by performing a following kernel equation:
\u2062
k
\u2061
(
x
,
y
)
=
exp
(
–
\u03c7
2
\u2061
(
x
,
y
)
2
\u2062
\u03c3
2
)
\u2062
\u2062
where
\u03c7
2
\u2061
(
x
,
y
)
\u2062
\u2062
is
\u2062
\u2062
defined
\u2062
\u2062
by
\u2062
\u2062
\u03c7
2
\u2061
(
S
,
M
)
=
\u2211
i
\u2062
(
S
i
–
M
j
\u2062
)
2
S
i
+
M
i
.
10. The method of claim 1, wherein in the obtaining of the similarity degree, the similarity degree is determined by comparing Euclidian distances.
11. The method of claim 1, wherein the target image is pre-registered before receipt of the query image.
12. An object verification apparatus comprising:
a target image Chi square inner product unit to obtain a target Chi square inner product based on a target local binary pattern (LBP) feature of a target image and a target kernel LBP feature to generate a target image feature vector;
a target image KFDA projection unit to project the target image feature vector to a kernel fisher discriminant analysis (KFDA) basis vector;
a query image Chi square inner product unit to obtain a query Chi square inner product based on a query LBP feature of a query image and a query kernel LBP feature to generate a query image feature vector;
a query image KFDA projection unit to project the query image feature vector to the KFDA basis vector; and
a similarity degree acquisition unit to compare the projected target image feature vector and the projected query image feature vector and provide a similarity degree between the target image and the query image and to output the similarity degree as a result of verification between the target image and the query image for object verification.
13. The apparatus of claim 12, wherein the target kernel LBP feature and query kernel LBP feature are a same kernel LBP feature.
14. The apparatus of claim 12, further comprising a basis vector generation unit to generate the KFDA basis vector by using at least one training LBP feature of at least one training image.
15. The apparatus of claim 14, wherein the basis vector generation unit comprises:
a training image Chi square inner product unit to obtain a training Chi square inner product based on at least one training LBP feature of at least one training image and a training kernel LBP feature; and
a KFDA basis vector generation unit to generate the KFDA basis vector by performing a KFDA for a result of the training image Chi square inner product unit.
16. The apparatus of claim 15, wherein the training kernel LBP feature, the target kernel LBP feature, and query kernel LBP feature are a same kernel LBP feature.
17. The apparatus of claim 15, wherein at least one of the training Chi square inner product, the target Chi square inner product, and the query Chi square inner product is obtained by performing a following kernel equation:
k
\u2061
(
x
,
y
)
=
exp
(
–
\u03c7
2
\u2061
(
x
,
y
)
2
\u2062
\u03c3
2
)
\u2062
\u2062
where
\u03c7
2
\u2061
(
x
,
y
)
\u2062
\u2062
is
\u2062
\u2062
defined
\u2062
\u2062
by
\u2062
\u2062
\u03c7
2
\u2061
(
S
,
M
)
=
\u2211
i
\u2062
(
S
i
–
M
j
\u2062
)
2
S
i
+
M
i
.
18. The apparatus of claim 15, wherein the at least one training LBP feature of the at least one training image is expressed as a corresponding LBP histogram by performing a corresponding LBP operation for the at least one training image.
19. The apparatus of claim 15, wherein the similarity degree of the similarity degree acquisition unit is determined by comparing Euclidian distances.
20. The apparatus of claim 12, wherein at least one of the target Chi square inner product and the query Chi square inner product is obtained by performing a following kernel equation:
k
\u2061
(
x
,
y
)
=
exp
(
–
\u03c7
2
\u2061
(
x
,
y
)
2
\u2062
\u03c3
2
)
\u2062
\u2062
where
\u03c7
2
\u2061
(
x
,
y
)
\u2062
\u2062
is
\u2062
\u2062
defined
\u2062
\u2062
by
\u2062
\u2062
\u03c7
2
\u2061
(
S
,
M
)
=
\u2211
i
\u2062
(
S
i
–
M
j
\u2062
)
2
S
i
+
M
i
.
21. The apparatus of claim 12, wherein the similarity degree of the similarity degree acquisition unit is determined by comparing Euclidian distances.
22. The apparatus of claim 12, wherein the apparatus is a face similarity determination apparatus and the similarity degree acquisition unit determines a similarity degree of a target face image and a query face image.
23. The apparatus of claim 12, wherein the target image is pre-registered before receipt of the query image.
24. An object verification method, comprising:
projecting a local binary pattern (LBP) based query image feature vector to a kernel fisher discriminant analysis (KFDA) basis vector; and
providing a similarity degree between a target image and the query image based on a projected target image feature vector and the projected LBP based query image feature vector and outputting the similarity degree as a result of verification between the target image and the query image for object verification.
25. The method of claim 24, further comprising generating the LBP based query image feature vector through a Chi square square inner product of a query LBP feature of the query image and a query kernel LBP feature.
26. The method of claim 24, wherein the projected target image feature vector is a LBP target image feature vector generated through a Chi square square inner product of a target LBP feature of the target image and a target query kernel LBP feature.
27. A medium comprising computer readable code to implement the method of claim 1.
28. A medium comprising computer readable code to implement the method of claim 24.
29. The method of claim 1, wherein the method is a face similarity determination method and the similarity degree is a similarity degree of a target face image and a query face image.
The claims below are in addition to those above.
All refrences to claim(s) which appear below refer to the numbering after this setence.
1. A semiconductor device for use in physical quantity distribution detection comprising:
an effective area including in a unit component:
a charge generating portion which generates charge corresponding to an incident electromagnetic wave; and
a unit signal generating portion which generates an analog unit signal in accordance with the charge generated by the charge generating portion; and
a reference signal generating portion which generates a reference signal to convert the unit signal to digital data as a functional component to the unit signal to digital data;
a comparing portion which compares the unit signal with the reference signal generated by the reference signal generating portion; and
a counter portion which conducts a count process by a predetermined count clock in parallel with the comparison process in the comparing portion and holds a count value at a time when the comparison process in the comparing portion is finished,
wherein on a surface onto which the electromagnetic wave of the individual charge generating portions enters in the effective area, any one of color filters of a color separation filter formed of a combination of color filters of multiple colors to acquire color information is disposed, and
the reference signal generating portion has individual reference signal generating and outputting portions which generate and output the reference signal fewer than the number of color components of color filters in a repeat unit for color filter arrangement in a predetermined direction in accordance with a unit of readout and in a direction different from the predetermined direction in accordance with the unit of readout and by the number of the color filters in the repeat unit for color filter arrangement in the predetermined direction in accordance with the unit of readout for the unit signal,
wherein the individual reference signals independently outputted from the reference signal generating and outputting portions are basically directly transmitted to the comparing portion corresponding to the color filters having a common color property in a predetermined direction through a common signal line.
2. The semiconductor device according to claim 1, wherein the individual reference signal generating and outputting portions generate the reference signal which varies with a change property in accordance with a color property of the color filter disposed in the unit component of a processing target, and change the change property in accordance with changing a combination of colors forming a repeat unit for color filter arrangement, the combination is switched associated with switching the unit of readout for the unit signal.
3. The semiconductor device according to claim 1, wherein the individual reference signal generating and outputting portions have: individual color corresponding reference signal generating portions which generate and output the reference signal in accordance with the color property of the corresponding color filter by the number of the color filters in the repeat unit for color filter arrangement in a direction different from a predetermined direction in accordance with the unit of readout, and a selecting portion which selects any one of the individual reference signals independently outputted from the color corresponding reference signal generating portions in accordance with switching the unit of readout of a processing target and outputs it to a corresponding signal line.
4. The semiconductor device according to claim 3, wherein the color corresponding reference signal generating portion has a change property control portion which controls the reference signal outputted from the color corresponding reference signal generating portion so as to vary with a change property in accordance the color property of the corresponding color filter.
5. The semiconductor device according to claim 1, wherein the individual reference signal generating and outputting portions in common have: a first color corresponding reference signal generating portion which generates and outputs the reference signal in accordance with the color property of the corresponding color filter and is in common used with respect to multiple color components at every color component in the repeat unit for color filter arrangement in a predetermined direction in accordance with the unit of readout and in a direction different from the predetermined direction in accordance with the unit of readout, the individual reference signal generating and outputting portions separately have: a second color corresponding reference signal generating portion which generates and outputs the reference signal in accordance with the color property of the corresponding color filter for an independent color component in the repeat unit for the arrangement at every color component; and a selecting portion which selects any one of the reference signals independently outputted from the first color corresponding reference signal generating portion and the second color corresponding reference signal generating portion in accordance with switching the unit of readout of a processing target and outputs it to a corresponding signal line.
6. The semiconductor device according to claim 5, wherein each of the first and second color corresponding reference signal generating portions has a change property control portion which controls the reference signal outputted from the color corresponding reference signal generating portion so as to vary with a change property in accordance with the color property of the corresponding color filter.
7. The semiconductor device according to claim 1, wherein the individual reference signal generating and outputting portions have: a color corresponding reference signal generating portion which includes multiple constant current sources arranged in parallel and generates and outputs the reference signal; a constant current source selecting portion which selects the multiple constant current sources based on a predetermined control signal; and a change property control portion which controls the reference signal outputted from the color corresponding reference signal generating portion so as to vary with a change property in accordance with the color property of the corresponding color filter by controlling current carried through the multiple constant current sources.
8. The semiconductor device according to claim 7, wherein the change property control portion is configured to have a reference constant current source having a current mirror structure with respect to the multiple constant current sources and to adjust current carried through the reference constant current source.
9. The semiconductor device according to claim 7, wherein the color corresponding reference signal generating portions are disposed by the number of the color filters in a repeat unit for the color filter arrangement in at least any one of a predetermined direction in accordance with the unit of readout and a direction different from the predetermined direction in accordance with the unit of readout.
10. The semiconductor device according to claim 7, wherein the constant current source selecting portion is disposed for each of the individual reference signal generating and outputting portions.
11. The semiconductor device according to claim 7, wherein the constant current source selecting portion is in common disposed with respect to the individual reference signal generating and outputting portions.
12. The semiconductor device according to claim 1, wherein the individual reference signal generating and outputting portions have: an initial value setting portion which sets an initial value based on a viewpoint different from the color property of the color filter disposed in the unit component of a processing target.
13. The semiconductor device according to claim 12, wherein on the side onto which the electromagnetic wave of the individual charge generating portions enters in a reference area except the effective area, it is configured that the unit signal outputted from the unit signal generating portion in the reference area provides a reference level to the unit signal outputted from the unit signal generating portion in the effective area, and the initial value setting portion sets the initial value defined by a viewpoint different from the color property based on the reference level.
14. The semiconductor device according to claim 12, wherein the initial value setting portion sets the initial value defined by a viewpoint different from the color property based on an offset component which is caused by a circuit forming the unit component and the AD converting portion and contained in the unit signal outputted from the unit signal generating portion in the effective area.
15. The semiconductor device according to claim 12, wherein on the side onto which the electromagnetic wave of the individual charge generating portions enters in a reference area except the effective area, it is configured that the unit signal outputted from the unit signal generating portion in the reference area provides a reference level to the unit signal outputted from the unit signal generating portion in the effective area, and the initial value setting portion sets the initial value defined by a viewpoint different from the color property based on an offset component which is caused by a circuit forming the unit component and the AD converting portion and contained in the unit signal outputted from the unit signal generating portion in the effective area.
16. The semiconductor device according to claim 12, wherein the individual reference signal generating and outputting portions have a color corresponding reference signal generating portion which includes multiple constant current sources arranged in parallel and generates and outputs the reference signal, and the initial value setting portion is arranged in parallel with the multiple constant current sources and controls the reference signal outputted from the color corresponding reference signal generating portion so as to vary with the initial value based on a viewpoint different from the color property of the color filter disposed in the unit component of a processing target.
17. The semiconductor device according to claim 16, wherein the initial value setting portion has an initial value setting current source which superimposes current providing the initial value on current carried through the multiple constant current sources arranged in parallel, wherein current carried through the initial value setting current source is configured to be adjustable.
18. The semiconductor device according to claim 1, wherein the counter portion is configured to allow selecting any one of a down count mode and an up count mode to conduct a count process.
19. The semiconductor device according to claim 18, wherein the unit signal is indicated as containing a reference component and a signal component, and the counter portion switches the modes of the count process in accordance with whether the comparing portion is doing the comparison process for any one of the reference component and the signal component.
20. A solid-state imaging device comprising:
a color filter portion having a plurality of color filters for acquiring color information of light from an object incident on said color filter portion;
an image sensing portion including pixels that receives the light through said color filter portion;
AD conversion portion that converts image signals transferred from said image sensing portion to digital signals; and
a reference signal generating portion that supplies reference signals for said AD conversion portion;
wherein the reference signal generating portion has individual reference signal generating and outputting portions which generate and output the reference signals fewer than the number of color components of said plurality of color filters in a repeat unit for color filter arrangement in a predetermined direction in accordance with a unit of readout and in a direction different from the predetermined direction in accordance with the unit of readout and by the number of the plurality of color filters in the repeat unit for color filter arrangement in the predetermined direction in accordance with the unit of readout for the unit signal, the individual reference signals independently outputted from the reference signal generating portion are basically directly transmitted to a comparing portion corresponding to the plurality of color filters having a common color property in a predetermined direction through a common signal line.