US 10,795,047 B2Granted

Methods for removing a background object from an image

AssigneeVoti Inc.

InventorsEmmanuel St-Aubin, Philippe Desjeans-Gauthier, Ola El Bakry, Simon Archambault, William Awad

Life of this patent

Estimating the term…

Loading patent…

Abstract

There is provided a method for assigning an attribute to an unknown object overlapping with a predetermined background object. The unknown object is scanned overlapping with the background object within an x-ray scanning device to obtain a plurality of dual-energy attenuation images having attenuation information representing the background object and an overlap region wherein the background object and the unknown object overlap. The dual-energy attenuation images are decomposed into reference material equivalent path length images. The reference material equivalent path lengths representing the background object in the overlap region are determined and eliminated from the overlap region to provide reference material equivalent path length images having first and second reference material equivalent path lengths through only the unknown object.

Claims61

Dependent claims

A method for assigning an attribute to an unknown object overlapping with a predetermined background object comprising the steps of:

scanning the predetermined background object in a plurality of positions and orientations within an x-ray scanning device to obtain a first plurality of predetermined background object dual-energy attenuation images each having predetermined background object dual-energy attenuation information representing the predetermined background object;
decomposing the predetermined background object dual-energy attenuation images into predetermined background object dual-reference material equivalent path length images having predetermined background object reference material equivalent path lengths passing through the predetermined background object;
scanning the unknown object at least partially overlapping with the predetermined background object within the x-ray scanning device to obtain a second plurality of dual-energy attenuation images having dual-energy attenuation information representing the background object and an overlap region wherein the background object and the unknown object overlap;
decomposing the second plurality of dual-energy attenuation images into reference material equivalent path length images wherein the overlap region has first and second reference material equivalent path lengths passing through both the background object and the unknown object;
determining the position and orientation of the background object in one of the second plurality of dual-energy attenuation images and the second reference material equivalent path length images of the unknown object by using a segmentation algorithm to localize the background object;
determining, by comparison, corresponding ones of the plurality of predetermined background object reference material equivalent path length images which most closely corresponds with the position and orientation of the background object in the dual-reference material equivalent path length images of the unknown object;
and,
eliminating the predetermined background object reference material equivalent path lengths of the corresponding ones of the plurality of the predetermined background object reference material equivalent path length images from the overlap region in the dual-reference material equivalent path length images of the unknown object to provide reference material equivalent path length images having first and second reference material equivalent path lengths through only the unknown object.

5 more elements

A method for assigning an attribute to an unknown object overlapping with an unknown background object having homogenous composition and thickness comprising the steps of:

scanning the unknown object at least partially overlapping with a portion of the background object within the x-ray scanning device to obtain a plurality of dual-energy attenuation images each having pixels distributed in rows and columns and each having dual-energy attenuation information;
decomposing each of the dual-energy attenuation images into dual-reference material equivalent path length images including a background region with first and second reference material equivalent path lengths passing through only the background object and an overlap region with first and second reference material equivalent path lengths passing through the unknown object overlapping with the background object;
determining the background region and the overlap region by using a segmentation algorithm to localize the background region;
determining one of an average, a median and a mean of the first and second reference material equivalent path lengths passing through only the background object in each column;
eliminating the one of the average, the median and the mean of the first and second reference material equivalent path lengths passing through only the background object from the first and second reference material equivalent path lengths of each column of the overlap region to determine first and second reference material equivalent path lengths representing only the unknown object.

2 more elements

A method for assigning an attribute to an unknown object overlapping with a background object having homogenous composition and non-uniform known thickness comprising the steps of:

scanning the unknown object at least partially overlapping with the background object within the x-ray scanning device to obtain dual-energy attenuation images each having pixels distributed in rows and columns and each having dual-energy attenuation information;
decomposing each of the dual-energy attenuation images into dual-reference material equivalent path length images including a background region with first and second reference material equivalent path lengths passing through only the background object and an overlap region with first and second reference material equivalent path lengths passing through the unknown object overlapping with the background object;
determining the effective atomic number of each pixel of the dual-reference material equivalent path length images;
determining the mass thickness of each pixel of the background region and the overlap region in the dual-reference material equivalent path length images;
localizing the background region and the overlap region in one of the dual-energy attenuation images and the dual-reference material equivalent path length images by using a segmentation algorithm;
obtaining the known background path lengths passing through the background region;
dividing the mass thickness of each pixel of the background region by the known background path lengths to determine the mass density of each pixel of the background region;
determining one of an average, a median and a mean of the effective atomic number across all pixels of the background region;
determining one of an average, a median and a mean of the mass density across all pixels of the background region;
for every pixel in the overlap region, multiplying the mass density of the background region with the background object path lengths;
for every pixel in the overlap region, eliminating the mass thickness of the background region from the total mass thickness of the background region and the overlap region to provide an unknown object mass thickness;
for every pixel in the overlap region, dividing the unknown object mass thickness by the path length through the unknown object of interest to provide the density of the unknown object;
calculating the effective atomic number of the unknown object by isolating the unknown object effective atomic number using the following equation which links the total mass thickness and effective atomic number values of every pixel with those of the object and background in the overlap region:
g[Zeff(i,j)]ρt(i,j)=g[Zo]ρoto(i,j)+g[Zb]ρbtb(i,j)
where Z is the effective atomic number of every pixel representing the unknown object and background object in the overlap region, where g[Z] is an invertible function of Z, where o and b represent the unknown object and background object respectively, where ρt is mass thickness, and where i and j represent the pixel of the ith row and jth column;
and isolating the effective atomic number of the unknown object Zo gives:
Z
o
⁡
(
i
,
j
)
=
g
-
1
⁢
{
g
⁡
[
z
eff
⁡
(
i
,
j
)
]
⁢
ρ
⁢
⁢
t
⁡
(
i
,
j
)
-
g
⁡
[
Z
_
b
]
⁢
ρ
⁢
⁢
t
o
∈
ob
⁡
(
i
,
j
)
ρ
⁢
⁢
t
o
∈
ob
⁡
(
i
,
j
)
}
;
and,
determining dual-reference material equivalent path lengths of the unknown object t1o(i,j) and t2o(i,j) by solving the following system of equation for every pixel (i,j) in the overlap region ob:
{
ρ
⁢
⁢
t
o
∈
ob
⁡
(
i
,
j
)
=
ρ
1
⁢
t
1
o
⁡
(
i
,
j
)
+
ρ
2
⁢
t
2
o
⁡
(
i
,
j
)
g
⁡
[
Z
o
⁡
(
i
,
j
)
]
⁢
ρ
⁢
⁢
t
o
∈
ob
⁡
(
i
,
j
)
=
g
⁡
[
Z
1
]
⁢
ρ
1
⁢
t
1
o
⁡
(
i
,
j
)
+
g
⁡
[
Z
2
]
⁢
ρ
2
⁢
t
2
o
⁡
(
i
,
j
)
wherein the solutions are:
t
1
o
⁡
(
i
,
j
)
=
ρ
⁢
⁢
t
o
∈
ob
⁡
(
i
,
j
)
ρ
1
⁢
{
g
⁡
[
Z
o
⁡
(
i
,
j
)
]
-
g
⁡
[
Z
2
]
g
⁡
[
Z
1
]
-
g
⁡
[
Z
2
]
}
t
2
o
⁡
(
i
,
j
)
=
ρ
⁢
⁢
t
o
∈
ob
⁡
(
i
,
j
)
ρ
2
⁢
{
g
⁡
[
Z
1
]
-
g
⁡
[
Z
o
⁡
(
i
,
j
)
]
g
⁡
[
Z
1
]
-
g
⁡
[
Z
2
]
}
.

317 more elements

Dependencies are inferred from the claim text.

Facts

Application: 16/229,860Filed 21 Dec 2018 · us provisional application
Priority: 9 Jan 2018
Published: US 2019/0213740 A111 Jul 2019
Granted: 6 Oct 2020
Term ends: Not estimatedThe source records don’t support an estimate.

Classification30 codes

G01V 5/224
Active interrogation, i.e. by irradiating objects or goods using external radiation sources, e.g. using gamma rays or cosmic rays: multiple energy…
G01N 23/04
By transmitting the radiation through the material: and forming images of the material
G01N 23/083
And measuring the absorption: the radiation being X-rays
G01N 23/087
The radiation being X-rays: using polyenergetic X-rays