Method and apparatus for displacement determination by motion compensation
View Patent ↗Motion estimator apparatus and methods are presented in which a fully constrained nonlinear system of equations combing forward and backward displaced frame difference (DFD) equations with a plurality of displacement vector invariant (DVI) equations is solved using the input data from two image frames without approximation and without imposing any additional constraints or assumptions to obtain an estimated displacement field. Also presented is an adaptive framework for solving a system of motion estimation equations.
1. A method for processing an image sequence, the method comprising:
providing a fully constrained nonlinear equation set including:
a forward displaced frame difference equation,
a backward displaced frame difference equation, and
a plurality of displacement vector invariant equations;
receiving an input image sequence comprising a pair of image frames individually including multidimensional image data corresponding to a plurality of pixel locations at different times; and
using at least one processor, solving the equation set using an iteration equation and the image data of the pair of image frames to determine a displacement field describing displacement vectors at pixel locations at one time by solving displacement field values at another time from the plurality of displacement vector invariant equations,
wherein the multidimensional image data of the pair of frames is two-dimensional;
wherein the forward displaced frame difference equation is l(i+Δx ij (t 1 ), j+Δy ij (t 1 ), t 2 )−l ij (t 1 )=0;
wherein the backward displaced frame difference equation is l ij (t 2 )−l(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 ); and
wherein the plurality of displacement vector invariant equations consists of:
a first forward displacement vector invariant equation, being Δx ij (t 2 )+Δx(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0, and
a second forward displacement vector invariant equation, being Δy ij (t 2 )+Δy(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0;
where l is an intensity of the image data; i is a pixel index in a horizontal direction x; j is a pixel index in a vertical direction y orthogonal to the horizontal direction x; t 1 is the one frame time; t 2 is the other frame time; Δx ij (t 1 ) is a displacement vector at pixel location i and j at the one time in the horizontal direction; Δx ij (t 2 ) is a displacement vector at pixel location i and j at the other time in the horizontal direction; Δy ij (t 1 ) is a displacement vector at pixel location i and j at the one time in the vertical direction; and Δy ij (t 2 ) is a displacement vector at pixel location i and j at the other time in the vertical direction.
2. The method of claim 1 :
wherein solving the equation set comprises solving the forward displacement vector invariant equations in the equation set using a lookup table of records at a time t that are off indexed pixel locations, with each record including position coordinates x=i+Δx ij (t 1 ) and y=j+Δy ij (t 1 ), and matrix values of Δx ij (t 1 ) and Δy ij (t 1 ) for all iε[0,N x −1]∩jε[0,N y −1]; and
wherein solving the equation set comprises, for all pixel locations, evaluating all the values of the matrix field Δx ij (t 2 ) and Δy ij (t 2 ) on a given pixel location by a bilinear interpolation method and a searching algorithm that finds three or four available values of the field Δx(x, y, t 2 ) and Δy(x, y, t 2 ) in the lookup table for all nearest points around the given pixel location.
3. The method of claim 1 , wherein solving the equation set comprises solving the forward displacement vector invariant equations in the equation set by a damped Newton-Raphson method using iteration equations for solving the matrix field at time t 2 for all pixel locations i and j, the iteration equations being given by:
(
Δ
x
ij
(
m
+
1
)
(
t
2
)
Δ
y
ij
(
m
+
1
)
(
t
2
)
)
=
(
Δ
x
ij
(
m
)
(
t
2
)
Δ
y
ij
(
m
)
(
t
2
)
)
-
(
∂
ɛ
1
(
m
)
∂
Δ
x
ij
(
m
)
(
t
2
)
∂
ɛ
1
(
m
)
∂
Δ
y
ij
(
m
)
(
t
2
)
∂
ɛ
2
(
m
)
∂
Δ
x
ij
(
m
)
(
t
2
)
∂
ɛ
2
(
m
)
∂
Δ
y
ij
(
m
)
(
t
2
)
)
-
1
(
ɛ
1
(
m
)
ɛ
2
(
m
)
)
where m is an iteration index; and
wherein solving the equation set comprises solving evaluating derivatives with respect variables Δx ij (t 2 ) and Δy ij (t 2 ) in the iteration equations by the following bilinear function:
{
Δ
x
(
x
,
y
,
t
1
)
,
Δ
y
(
x
,
y
,
t
1
)
}
=
{
Δ
x
pq
(
t
1
)
,
Δ
y
pq
(
t
1
)
}
(
n
x
-
x
+
p
)
(
n
y
-
y
+
q
)
n
x
n
y
++
{
Δ
x
p
+
n
x
,
q
(
t
1
)
,
Δ
y
p
+
n
x
,
q
(
t
1
)
}
(
x
-
p
)
(
n
y
-
y
+
q
)
n
x
n
y
++
{
Δ
x
p
,
q
+
n
y
(
t
1
)
,
Δ
y
p
,
q
+
n
y
(
t
1
)
}
(
n
x
-
x
+
p
)
(
y
-
q
)
n
x
n
y
++
{
Δ
x
p
+
n
x
,
q
+
n
y
(
t
1
)
,
Δ
y
p
+
n
x
,
q
+
n
y
(
t
1
)
}
(
x
-
p
)
(
y
-
q
)
n
x
n
y
and
{
p
,
q
}
=
{
n
x
⌊
x
n
x
⌋
,
n
y
⌊
y
n
y
⌋
}
,
where └ ┘ denotes an integer operator, and n x and n y are a number of interpolation points on x and y directions.
4. A method for processing an image sequence, the method comprising:
providing a fully constrained nonlinear equation set including:
a forward displaced frame difference equation,
a backward displaced frame difference equation, and
a plurality of displacement vector invariant equations;
receiving an input image sequence comprising a pair of image frames individually including multidimensional image data corresponding to a plurality of pixel locations at different times; and
using at least one processor, solving the equation set using an iteration equation and the image data of the pair of image frames to determine a displacement field describing displacement vectors at pixel locations at one time by solving displacement field values at another time from the plurality of displacement vector invariant equations,
wherein the multidimensional image data of the pair of frames is two-dimensional;
wherein the forward displaced frame difference equation is l(i+Δx ij (t 1 ), j+Δy ij (t 1 ), t 2 )−l ij (t 1 )=0;
wherein the backward displaced frame difference equation is l ij (t 2 )−l(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0; and
wherein the plurality of displacement vector invariant equations consists of:
a first backward displacement vector invariant equation, being Δx(x,y,t 1 )+Δx(x+Δx(x,y,t 1 ),y+Δy(x,y,t 1 ),t 2 )=0, and
a second backward displacement vector invariant equation, being Δy(x,y,t 1 )+Δy(x+Δx(x,y,t 1 ),y+Δy(x,y,t 1 ),t 2 )=0;
where l is an intensity of the image data; i is a pixel index in a horizontal direction x; j is a pixel index in a vertical direction y orthogonal to the horizontal direction x; t 1 is one of the different times; t 2 is another of the different times; Δx ij (t 1 ) is a displacement vector at pixel location ij at the one time in the horizontal direction; Δx ij (t 2 ) is a displacement vector at pixel location ij at the other time in the horizontal direction; Δy ij (t 1 ) is a displacement vector at pixel location ij at the one time in the vertical direction; and Δy ij (t 2 ) is a displacement vector at pixel location ij at the other time in the vertical direction.
5. The method of claim 4 :
wherein solving the equation set comprises solving the backward displacement vector invariant equations in the equation set using a lookup table of records at a time t that are off indexed pixel locations, with each record including position coordinates x=i+Δx ij (t 1 ) and y=j+Δy ij (t 1 ), and matrix values of Δx ij (t 1 ) and Δy ij (t 1 ) for all iε[0,N x −1]∩jε[0,N y −1]; and
wherein solving the equation set comprises, for all pixel locations, evaluating all the values of the matrix field Δx ij (t 2 ) and Δy ij (t 2 ) on a given pixel location by a bilinear interpolation method and a searching algorithm that finds three or four available values of the field Δx(x, y, t 2 ) and Δy(x, y, t 2 ) in the lookup table for all nearest points around the given pixel location.
6. The method of claim 4 , wherein solving the equation set comprises solving the backward displacement vector invariant equations in the equation set by a damped Newton-Raphson method using iteration equations for solving the matrix field at time t 1 to solve for X 1 and Y 1 at time t for all pixel locations i and j, the iteration equations being given by:
(
x
1
(
m
+
1
)
y
1
(
m
+
1
)
)
=
(
x
1
(
m
)
y
1
(
m
)
)
-
(
∂
e
1
(
m
)
∂
x
1
(
m
)
∂
e
1
(
m
)
∂
y
1
(
m
)
∂
e
2
(
m
)
∂
x
1
(
m
)
∂
e
2
(
m
)
∂
y
1
(
m
)
)
-
1
(
e
1
(
m
)
e
2
(
m
)
)
where m is an iteration index, and where variables e 1 and e 2 are defined as follows:
{
e
1
=
x
1
+
Δ
x
(
x
1
,
y
1
,
t
1
)
-
i
=
0
e
2
=
y
1
+
Δ
y
(
x
1
,
y
1
,
t
1
)
-
j
=
0
;
and
wherein solving the equation set comprises solving the backward displacement vector invariant equations for all pixel locations using the following:
{
Δ
x
ij
(
t
2
)
=
-
Δ
x
(
x
1
,
y
1
,
t
1
)
Δ
y
ij
(
t
2
)
=
-
Δ
y
(
x
1
,
y
1
,
t
1
)
.
7. The method of claim 1 , wherein solving the equation set comprises using at least one iteration equation derived from the equation set by conversion to a fully or over-constrained system using a nonlinear least squares model of the displacement field.
8. The method of claim 1 , wherein solving the equation set comprises using at least one bilinear polynomial function expressing a multidimensional displacement field.
9. A motion estimator apparatus, comprising:
at least one processor; and
a memory storing a fully constrained nonlinear equation set including:
a forward displaced frame difference equation,
a backward displaced frame difference equation, and
a plurality of displacement vector invariant equations;
the at least one processor operative to receive an input image sequence comprising a pair of image frames individually including multidimensional image data corresponding to a plurality of pixel locations at different times; and
the at least one processor operative to solve the equation set using an iteration equation and the image data of the pair of image frames to determine a displacement field describing displacement vectors at pixel locations at one time by solving displacement field values at another time from the plurality of displacement vector invariant equations,
wherein the multidimensional image data of the pair of frames is two-dimensional;
wherein the forward displaced frame difference equation is l(i+Δx ij (t 1 ), j+Δy ij (t 1 ), t 2 )−l ij (t 1 )=0;
wherein the backward displaced frame difference equation is l ij (t 2 )−l(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0; and
wherein the plurality of displacement vector invariant equations consists of:
a first forward displacement vector invariant equation, being Δx ij (t 2 )+Δx(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0, and
a second forward displacement vector invariant equation, being Δy ij (t 2 )+Δy(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0;
where l is an intensity of the image data; i is a pixel index in a horizontal direction x; j is a pixel index in a vertical direction y orthogonal to the horizontal direction x; t 1 is the one time; t 2 is the other time; Δx ij (t 1 ) is a displacement vector at pixel location i and j at the one time in the horizontal direction; Δx ij (t 2 ) is a displacement vector at pixel location i and j at the other time in the horizontal direction; Δy ij (t 1 ) is a displacement vector at pixel location i and j at the one time in the vertical direction; and Δy ij (t 2 ) is a displacement vector at pixel location i and j at the other time in the vertical direction.
10. The motion estimator apparatus of claim 9 :
wherein the at least one processor is operative to solve the forward displacement vector invariant equations in the equation set using a lookup table of records at a time t that are off indexed pixel locations, with each record including position coordinates x=i+Δx ij (t 1 ) and y=j+Δy ij (t 1 ), and matrix values of Δx ij (t 1 ) and Δy ij (t 1 ) for all iε[0,N x −1]∩jε[0,N y −1]; and
wherein the at least one processor is operative, for all pixel locations, to evaluate all the values of the matrix field Δx ij (t 2 ) and Δy ij (t 2 ) on a given pixel location by a bilinear interpolation method and a searching algorithm that finds three or four available values of the field Δx(x, y, t 2 ) and Δy(x, y, t 2 ) in the lookup table for all nearest points around the given pixel location.
11. The motion estimator apparatus of claim 9 , wherein the at least one processor is operative to solve the forward displacement vector invariant equations in the equation set by a damped Newton-Raphson method using iteration equations for solving the matrix field at time t 2 for all pixel locations i and j, the iteration equations being given by:
(
Δ
x
ij
(
m
+
1
)
(
t
2
)
Δ
y
ij
(
m
+
1
)
(
t
2
)
)
=
(
Δ
x
ij
(
m
)
(
t
2
)
Δ
y
ij
(
m
)
(
t
2
)
)
-
(
∂
ɛ
1
(
m
)
∂
Δ
x
ij
(
m
)
(
t
2
)
∂
ɛ
1
(
m
)
∂
Δ
y
ij
(
m
)
(
t
2
)
∂
ɛ
2
(
m
)
∂
Δ
x
ij
(
m
)
(
t
2
)
∂
ɛ
2
(
m
)
∂
Δ
y
ij
(
m
)
(
t
2
)
)
-
1
(
ɛ
1
(
m
)
ɛ
2
(
m
)
)
where m is an iteration index; and
wherein the at least one processor is operative to evaluate derivatives with respect variables Δx ij (t 2 ) and Δy ij (t 2 ) in the iteration equations by the following bilinear function:
{
Δ
x
(
x
,
y
,
t
1
)
,
Δ
y
(
x
,
y
,
t
1
)
}
=
{
Δ
x
pq
(
t
1
)
,
Δ
y
pq
(
t
1
)
}
(
n
x
-
x
+
p
)
(
n
y
-
y
+
q
)
n
x
n
y
++
{
Δ
x
p
+
n
x
,
q
(
t
1
)
,
Δ
y
p
+
n
x
,
q
(
t
1
)
}
(
x
-
p
)
(
n
y
-
y
+
q
)
n
x
n
y
++
{
Δ
x
p
,
q
+
n
y
(
t
1
)
,
Δ
y
p
,
q
+
n
y
(
t
1
)
}
(
n
x
-
x
+
p
)
(
y
-
q
)
n
x
n
y
++
{
Δ
x
p
+
n
x
,
q
+
n
y
(
t
1
)
,
Δ
y
p
+
n
x
,
q
+
n
y
(
t
1
)
}
(
x
-
p
)
(
y
-
q
)
n
x
n
y
and
{
p
,
q
}
=
{
n
x
⌊
x
n
x
⌋
,
n
y
⌊
y
n
y
⌋
}
,
where └ ┘ denotes an integer operator, and n x and n y are a number of interpolation points on x and y directions.
12. A motion estimator apparatus, comprising:
at least one processor; and
a memory storing a fully constrained nonlinear equation set including:
a forward displaced frame difference equation,
a backward displaced frame difference equation, and
a plurality of displacement vector invariant equations;
the at least one processor operative to receive an input image sequence comprising a pair of image frames individually including multidimensional image data corresponding to a plurality of pixel locations at different times; and
the at least one processor operative to solve the equation set using an iteration equation and the image data of the pair of image frames to determine a displacement field describing displacement vectors at pixel locations at one time by solving displacement field values at another time from the plurality of displacement vector invariant equations,
wherein the multidimensional image data of the pair of frames is two-dimensional;
wherein the forward displaced frame difference equation is l(i+Δx ij (t 1 ), j+Δy ij (t 1 ), t 2 )−l ij (t 1 )=0;
wherein the backward displaced frame difference equation is l ij (t 2 )−l(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0; and
wherein the plurality of displacement vector invariant equations consists of:
a first backward displacement vector invariant equation, being Δx(x,y,t 1 )+Δx(x+Δx(x,y,t 1 ),y+Δy(x,y,t 1 ),t 2 )=0, and
a second backward displacement vector invariant equation, being Δy(x,y,t 1 )+Δy(x+Δx(x,y,t 1 ),y+Δy(x,y,t 1 ),t 2 )=0;
where l is an intensity of the image data; i is a pixel index in a horizontal direction x; j is a pixel index in a vertical direction y orthogonal to the horizontal direction x; t 1 is one of the different times; t 2 is another of the different times; Δx ij (t 1 ) is a displacement vector at pixel location ij at the one time in the horizontal direction; Δx ij (t 2 ) is a displacement vector at pixel location ij at the other time in the horizontal direction; Δy ij (t 1 ) is a displacement vector at pixel location ij at the one time in the vertical direction; and Δy ij (t 2 ) is a displacement vector at pixel location ij at the other time in the vertical direction.
13. The motion estimator apparatus of claim 12 :
wherein the at least one processor is operative to solve the backward displacement vector invariant equations in the equation set using a lookup table of records at a time t that are off indexed pixel locations, with each record including position coordinates x=i+Δx ij (t 1 ) and y=j+Δy ij (t 1 ), and matrix values of Δx ij (t 1 ) and Δy ij (t 1 ) for all iε[0,N x −1]∩jε[0,N y −1]; and
wherein the at least one processor is operative, for all pixel locations, evaluating all the values of the matrix field Δx ij (t 2 ) and Δy ij (t 2 ) on a given pixel location by a bilinear interpolation method and a searching algorithm that finds three or four available values of the field Δx(x, y, t 2 ) and Δy(x, y, t 2 ) in the lookup table for all nearest points around the given pixel location.
14. The motion estimator apparatus of claim 12 , wherein the at least one processor is operative to solve the backward displacement vector invariant equations in the equation set by a damped Newton-Raphson method using iteration equations for solving the matrix field at time t 1 to solve for X 1 and Y 1 at time t for all pixel locations i and j, the iteration equations being given by:
(
x
1
(
m
+
1
)
y
1
(
m
+
1
)
)
=
(
x
1
(
m
)
y
1
(
m
)
)
-
(
∂
e
1
(
m
)
∂
x
1
(
m
)
∂
e
1
(
m
)
∂
y
1
(
m
)
∂
e
2
(
m
)
∂
x
1
(
m
)
∂
e
2
(
m
)
∂
y
1
(
m
)
)
-
1
(
e
1
(
m
)
e
2
(
m
)
)
where m is an iteration index, and where variables e 1 and e 2 are defined as follows:
{
e
1
=
x
1
+
Δ
x
(
x
1
,
y
1
,
t
1
)
-
i
=
0
e
2
=
y
1
+
Δ
y
(
x
1
,
y
1
,
t
1
)
-
j
=
0
;
and
wherein the at least one processor is operative to solve the backward displacement vector invariant equations for all pixel locations using the following:
{
Δ
x
ij
(
t
2
)
=
-
Δ
x
(
x
1
,
y
1
,
t
1
)
Δ
y
ij
(
t
2
)
=
-
Δ
y
(
x
1
,
y
1
,
t
1
)
.
15. The motion estimator apparatus of claim 9 , wherein the at least one processor is operative to solve the equation set using at least one iteration equation derived from the equation set by conversion to a fully or over-constrained system using a nonlinear least squares model of the displacement field.
16. The motion estimator apparatus of claim 9 , wherein the at least one processor is operative to solve the equation set using at least one bilinear polynomial function expressing a multidimensional displacement field.
17. A non-transitory computer readable medium with computer executable instructions for:
providing a fully constrained nonlinear equation set including:
a forward displaced frame difference equation,
a backward displaced frame difference equation, and
a plurality of displaced vector invariant equations;
receiving an input image sequence comprising a pair of image frames individually including multidimensional image data corresponding to a plurality of pixel locations at different times; and
solving the equation set using an iteration equation and the image data of the pair of image frames to determine a displacement field describing displacement vectors at pixel locations at one time solving displacement field values at another time from the plurality of displacement vector invariant equations,
wherein the multidimensional image data of the pair of frames is two-dimensional;
wherein the forward displaced frame difference equation is l(i+Δx ij (t 1 ), j+Δy ij (t 1 ), t 2 )−l ij (t 1 )=0;
wherein the backward displaced frame difference equation is l ij (t 2 )−l(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0; and
wherein the plurality of displacement vector invariant equations consists of:
a first forward displacement vector invariant equation, being Δx ij (t 2 )+Δx(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0, and
a second forward displacement vector invariant equation, being Δy ij (t 2 )+Δy(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0;
where l is an intensity of the image data; i is a pixel index in a horizontal direction x; j is a pixel index in a vertical direction y orthogonal to the horizontal direction x; t 1 is the one time; t 2 is the other time; Δx ij (t 1 ) is a displacement vector at pixel location i and j at the one time in the horizontal direction; Δx ij (t 2 ) is a displacement vector at pixel location i and j at the other time in the horizontal direction; Δy ij (t 1 ) is a displacement vector at pixel location i and j at the one time in the vertical direction; and Δy ij (t 2 ) is a displacement vector at pixel location i and j at the other time in the vertical direction.
18. The non-transitory computer readable medium of claim 17 , comprising computer executable instructions for solving the forward displacement vector invariant equations in the equation set by a damped Newton-Raphson method using iteration equations for solving the matrix field at time t 2 for all pixel locations i and j, the iteration equations being given by:
(
Δ
x
ij
(
m
+
1
)
(
t
2
)
Δ
y
ij
(
m
+
1
)
(
t
2
)
)
=
(
Δ
x
ij
(
m
)
(
t
2
)
Δ
y
ij
(
m
)
(
t
2
)
)
-
(
∂
ɛ
1
(
m
)
∂
Δ
x
ij
(
m
)
(
t
2
)
∂
ɛ
1
(
m
)
∂
Δ
y
ij
(
m
)
(
t
2
)
∂
ɛ
2
(
m
)
∂
Δ
x
ij
(
m
)
(
t
2
)
∂
ɛ
2
(
m
)
∂
Δ
y
ij
(
m
)
(
t
2
)
)
-
1
(
ɛ
1
(
m
)
ɛ
2
(
m
)
)
where m is an iteration index; and
computer-executable instructions for solving evaluating derivatives with respect variables Δx ij (t 2 ) and Δy ij (t 2 ) in the iteration equations by the following bilinear function:
{
Δ
x
(
x
,
y
,
t
1
)
,
Δ
y
(
x
,
y
,
t
1
)
}
=
{
Δ
x
pq
(
t
1
)
,
Δ
y
pq
(
t
1
)
}
(
n
x
-
x
+
p
)
(
n
y
-
y
+
q
)
n
x
n
y
++
{
Δ
x
p
+
n
x
,
q
(
t
1
)
,
Δ
y
p
+
n
x
,
q
(
t
1
)
}
(
x
-
p
)
(
n
y
-
y
+
q
)
n
x
n
y
++
{
Δ
x
p
,
q
+
n
y
(
t
1
)
,
Δ
y
p
,
q
+
n
y
(
t
1
)
}
(
n
x
-
x
+
p
)
(
y
-
q
)
n
x
n
y
++
{
Δ
x
p
+
n
x
,
q
+
n
y
(
t
1
)
,
Δ
y
p
+
n
x
,
q
+
n
y
(
t
1
)
}
(
x
-
p
)
(
y
-
q
)
n
x
n
y
and
{
p
,
q
}
=
{
n
x
⌊
x
n
x
⌋
,
n
y
⌊
y
n
y
⌋
}
,
where └ ┘ denotes an integer operator, and n x and n y are a number of interpolation points on x and y directions.
19. A non-transitory computer readable medium with computer executable instructions for:
providing a fully constrained nonlinear equation set including:
a forward displaced frame difference equation,
a backward displaced frame difference equation, and
a plurality of displaced vector invariant equations;
receiving an input image sequence comprising a pair of image frames individually including multidimensional image data corresponding to a plurality of pixel locations at different times; and
solving the equation set using an iteration equation and the image data of the pair of image frames to determine a displacement field describing displacement vectors at pixel locations at one time solving displacement field values at another time from the plurality of displacement vector invariant equations,
wherein the multidimensional image data of the pair of frames is two-dimensional;
wherein the forward displaced frame difference equation is l(i+Δx ij (t 1 ), j+Δy ij (t 1 ), t 2 )−l ij (t 1 )=0;
wherein the backward displaced frame difference equation is l ij (t 2 )−l(i+Δx ij (t 2 ), j+Δy ij (t 2 ), t 1 )=0; and
wherein the plurality of displacement vector invariant equations consists of:
a first backward displacement vector invariant equation, being Δx(x,y,t 1 )+Δx(x+Δx(x,y,t 1 ),y+Δy(x,y,t 1 ),t 2 )=0, and
a second backward displacement vector invariant equation, being Δy(x,y,t 1 )+Δy(x+Δx(x,y,t 1 ),y+Δy(x,y,t 1 ),t 2 )=0;
where l is an intensity of the image data; i is a pixel index in a horizontal direction x; j is a pixel index in a vertical direction y orthogonal to the horizontal direction x; t 1 is one of the different times; t 2 is another of the different times; Δx ij (t 1 ) is a displacement vector at pixel location i and j at the one time in the horizontal direction; Δx ij (t 2 ) is a displacement vector at pixel location i and j at the other time in the horizontal direction; Δy ij (t 1 ) is a displacement vector at pixel location i and j at the one time in the vertical direction; and Δy ij (t 2 ) is a displacement vector at pixel location i and j at the other time in the vertical direction.
20. The non-transitory computer readable medium of claim 19 , comprising computer-executable instructions for solving the backward displacement vector invariant equations in the equation set by a damped Newton-Raphson method using iteration equations for solving the matrix field at time t 1 to solve for X 1 and Y 1 at time t for all pixel locations i and j, the iteration equations being given by:
(
x
1
(
m
+
1
)
y
1
(
m
+
1
)
)
=
(
x
1
(
m
)
y
1
(
m
)
)
-
(
∂
e
1
(
m
)
∂
x
1
(
m
)
∂
e
1
(
m
)
∂
y
1
(
m
)
∂
e
2
(
m
)
∂
x
1
(
m
)
∂
e
2
(
m
)
∂
y
1
(
m
)
)
-
1
(
e
1
(
m
)
e
2
(
m
)
)
where m is an iteration index, and where variables e 1 and e 2 are defined as follows:
{
e
1
=
x
1
+
Δ
x
(
x
1
,
y
1
,
t
1
)
-
i
=
0
e
2
=
y
1
+
Δ
y
(
x
1
,
y
1
,
t
1
)
-
j
=
0
;
and
comprising computer executable instructions for solving the backward displacement vector invariant equations for all pixel locations using the following:
{
Δ
x
ij
(
t
2
)
=
-
Δ
x
(
x
1
,
y
1
,
t
1
)
Δ
y
ij
(
t
2
)
=
-
Δ
y
(
x
1
,
y
1
,
t
1
)
.
21. The non-transitory computer readable medium of claim 17 , comprising computer-executable instructions for solving the equation set using at least one iteration equation derived from the equation set by conversion to a fully or over-constrained system using a nonlinear least squares model of the displacement field.
22. The non-transitory computer readable medium of claim 17 , comprising computer-executable instructions for solving the equation set using at least one bilinear polynomial function expressing a multidimensional displacement field.