U.S. patent application number 13/534584 was filed with the patent office on 2014-01-02 for image enhancement methods.
This patent application is currently assigned to 3M Innovative Properties Company. The applicant listed for this patent is Gerald P. Cook, Anthony D. Jacques. Invention is credited to Gerald P. Cook, Anthony D. Jacques.
Application Number | 20140002872 13/534584 |
Document ID | / |
Family ID | 48672808 |
Filed Date | 2014-01-02 |
United States Patent
Application |
20140002872 |
Kind Code |
A1 |
Cook; Gerald P. ; et
al. |
January 2, 2014 |
IMAGE ENHANCEMENT METHODS
Abstract
Methods of image enhancement are disclosed. In one aspect, the
method of image enhancement is for use with an image capture
device, such as a security document reader, for the attenuation,
separation or reduction of reflections from objects, such as
security documents.
Inventors: |
Cook; Gerald P.; (Bromsash,
GB) ; Jacques; Anthony D.; (Hampshire, GB) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Cook; Gerald P.
Jacques; Anthony D. |
Bromsash
Hampshire |
|
GB
GB |
|
|
Assignee: |
3M Innovative Properties
Company
|
Family ID: |
48672808 |
Appl. No.: |
13/534584 |
Filed: |
June 27, 2012 |
Current U.S.
Class: |
358/504 ;
358/406 |
Current CPC
Class: |
G06T 5/50 20130101; G06T
2207/30176 20130101; G06T 2207/10008 20130101 |
Class at
Publication: |
358/504 ;
358/406 |
International
Class: |
H04N 1/409 20060101
H04N001/409 |
Claims
1. An image enhancement method for an image capture device
comprising: illuminating an object placed on, in or adjacent to the
image capture device and capturing an image of the object from a
first position to obtain a first set of raw pixel data;
illuminating the object placed on, in or adjacent to the image
capture device, capturing an image of the object from a second
position, to obtain a second set of raw pixel data, where each
pixel in the second set of raw pixel data corresponds to a pixel in
the first set of raw pixel data representing a point on the object;
calibrating each of the first and second sets of raw pixel data
using a first set of image calibration pixel data to create a first
set of image pixel data and using a second set of image calibration
pixel data to create a second set of pixel image data, wherein each
pixel in the second set of image calibration pixel data corresponds
to a pixel in the first set of image calibration pixel data, and
each pixel in the first and second sets of image calibration pixel
data corresponds to a pixel in each of the first and second sets of
raw pixel data respectively; for a first pixel in the first set of
image pixel data: i) calculating the modulus of the pixel intensity
of said pixel in the first set of image pixel data minus the pixel
intensity of the corresponding pixel in the second set of image
pixel data and comparing the modulus to a pre-determined threshold
value; ii) if the modulus is greater than the threshold value,
selecting the one of the pixel in the first set of image pixel data
and the pixel in the second set of image pixel data with the lowest
pixel intensity as the output pixel; iii) if the modulus is less
than the threshold value, determining the one of the corresponding
pixel in the first set of image calibration pixel data and the
corresponding pixel in the second set of image calibration pixel
data having the highest pixel intensity, and selecting the
corresponding pixel in either the first set of image pixel data or
the second set of image pixel data as the output pixel; and
repeating steps i), ii) and iii) for at least a second pixel in the
first set of image pixel data and forming a set of final pixel data
from the resulting output pixels.
2. The image enhancement method of claim 1, wherein the object is
illuminated with visible light, infra-red light or ultraviolet
light.
3. The image enhancement method of claim 1, wherein when the object
is illuminated with visible light, the object is illuminated with
white light.
4. The image enhancement method of claim 3, wherein the pixel
intensity includes balanced red-green-blue components.
5. The image enhancement method of claim 3, wherein the pixel
intensity includes un-balanced red-green-blue components.
6. The image enhancement method of claim 3, wherein the pixel
intensity includes a maximum red, green or blue component.
7. The image enhancement method of claim 1, further comprising: for
each pixel in the first and second sets of raw pixel data,
measuring the intensity of single color reflections, and for pixels
representing the same region of the object, selecting the pixel
with the brightest single color intensity; and including said pixel
in a second set of final image data.
8. The image enhancement method of claim 1, further comprising:
adjusting the first and second sets of image pixel data with a
gamma correction.
9. The image enhancement method of claim 1, wherein the image
enhancement comprises the attenuation, separation or removal of
reflections.
10. The image enhancement method of claim 1, wherein the image
enhancement comprises the attenuation, separation or removal of
specular reflections.
11. The image enhancement method of claim 1, further comprising:
for each of the first and second sets of raw pixel data,
compensating the intensity values of each pixel for ambient
light.
12. The image enhancement method of claim 10, further comprising:
creating a set of ambient pixel data by imaging the object under no
illumination other than ambient light; and subtracting the set of
ambient pixel data from each of the first and second sets of raw
pixel data.
13. The image enhancement method of claim 1, wherein the object is
a security document.
14. The image enhancement method of claim 13, wherein the object an
identity document or a fiduciary document.
15. The image enhancement method of claim 13, wherein the object is
a passport, an identification card, or a driver's license.
16. The image enhancement method of claim 13, wherein the image
capture device is a security document reader.
17. The image enhancement method of claim 1, wherein the first
position is different from the second position.
18. The image enhancement method of claim 17, wherein the first
position is at a first angle relative to the object, and wherein
the second position is at a second angle relative to the object.
Description
[0001] Security documents such as passports, identification cards,
national healthcare cards, driver's licenses, entry passes,
ownership certificates, financial instruments, and the like, are
often assigned to a particular person by personalization data.
Personalization data, often present as printed images, can include
photographs, signatures, fingerprints, personal alphanumeric
information, and barcodes, and allows human or electronic
verification that the person presenting the document for inspection
is the person to whom the document is assigned. There is widespread
concern that forgery techniques can be used to alter the
personalization data on such a document, thus allowing
non-authorized people to pass the inspection step and use the
document in a fraudulent manner.
[0002] A number of security features have been developed to help
authenticate the document of value, thus assisting in preventing
counterfeiters from altering, duplicating or simulating a document
of value. Some of these security features may include overt
security features or covert security features. Overt security
features are features that are easily viewable to the unaided eye,
such features may include holograms and other diffractive optically
variable images, embossed images, and color-shifting films. In
contrast, covert security features include images only visible
under certain conditions, such as inspection under light of a
certain wavelength, polarized light, or retroreflected light. One
example of a laminate that includes both overt and covert security
features is 3M.TM. Confirm.TM. Laminate with Floating Image
Technology, which is commercially available from 3M Company based
in St. Paul, Minn. This security laminate may be used with security
documents, such as identification cards, badges and driver
licenses, and assists in providing identification, authentication
and to help protect against counterfeiting, alteration,
duplication, and simulation. Another example of a laminate that
includes both overt and covert security features is illustrated in
U.S. Pat. Publication No. 2003/0170425 A1 "Security Laminate,"
(Mann et al.).
[0003] In recent years there has been widespread adoption of
automated reading of security documents at border entry points and
other situations where the identity of a document holder requires
verification. Automated reading ranges from an optical scan of
OCR-readable data to the interrogation of an RFID chip within a
passport or identification card, which may then involve further
checking by an operator or verification by an automated system such
as an e-passport gate as found in major airports. Data may also be
contained in a magnetic strip or transferred wirelessly depending
on the format of the document in which identity information is
contained.
[0004] Optical reading of a security document is typically carried
out with document readers using one or a combination of visible,
infrared and ultraviolet light, depending on the information being
retrieved. Often overt and covert optical security features, such
as those discussed above, are included within security documents to
allow the document itself to be authenticated as genuine. As
discussed, covert security features may only be visible under
certain illumination, such as infrared or ultraviolet light, or
may, such as with a hologram, provide variable information when
illuminated from different directions. In each case the security
document is typically read by placing the document on a glass
platen of a document reader, such that the information contained on
the portion of the document in contact with the platen is
illuminated from within the document reader. Light reflected by the
document is reflected back into the reader and processed to form an
image of the information (e.g. text or covert or overt security
features) required. The quality of the image captured is affected
greatly by the manner in which the document reflects the incident
light.
[0005] A variety of security readers are known in the art. For
example, U.S. Pat. No. 6,288,842, "Security Reader for Automatic
Detection of Tampering and Alteration, (Mann) discloses a security
reader for reading and processing information about security
laminates. One example of a passport reader is commercially
available from 3M Company based in St. Paul, Minn., as the 3M.TM.
Full Page Reader.
[0006] Image enhancement by removal of unwanted reflections in
image capture devices is disclosed in U.S. Pat. No. 7,136,537,
"Specular Reflection in Captured Images," (Pilu et al.). In order
to remove specular reflections, two images are taken, one
containing specular reflections and one where such reflections are
absent. These images are blended together to create an image with
reduced specular reflection, allowing underlying features to be
seen. The apparatus used to achieve this effect is provided with an
adjustor that is able to vary the amount of specular reflection
appearing in the final image. Images are taken with one or more
strobes or flashes from various directions relative to the object
being imaged, and relies on each image having an absence of glare
patches seen in another image. Such a method therefore takes into
account reflections generated by ambient light conditions, and is
not suitable for use in a document reader, for example, where
illumination is well controlled and reflection features are
generated by artefacts in the document being imaged, rather than
artefacts generated by variations in ambient illumination.
SUMMARY
[0007] One aspect of the present invention provides an image
enhancement method for an image capture device. This method
comprises: illuminating an object placed on, in or adjacent to the
image capture device and capturing an image of the object from a
first position to obtain a first set of raw pixel data;
illuminating the object placed on, in or adjacent to the image
capture device and capturing an image of the object from a second
position, to obtain a second set of raw pixel data, wherein each
pixel in the second set of raw pixel data corresponds to a pixel in
the first set of raw pixel data representing a point on the object;
calibrating each of the first and second sets of raw pixel data
using a set of image calibration pixel data to create a first set
of image pixel data and a second set of pixel image data; and
calculating a first set of final image data by: comparing the first
and second sets of image pixel data; for pixels representing the
same point on the object, selecting the pixel with the lowest pixel
intensity; and including said pixel in the first set of final image
data.
[0008] The above summary of the present invention is not intended
to describe each disclosed embodiment or every implementation of
the present invention. The Figures and the detail description,
which follow, more particularly exemplify illustrative
embodiments.
BRIEF DESCRIPTION OF THE DRAWINGS
[0009] The present invention will be further explained with
reference to the appended Figures, wherein like structure is
referred to by like numerals throughout the several views, and
wherein:
[0010] FIG. 1 is a schematic side view of a document reader in
which an embodiment of the method of the present invention is
carried out;
[0011] FIG. 2 is a schematic side view of one type of optical
defect in a security document giving rise to a specular
reflection;
[0012] FIG. 3A is a schematic illustration of an image of a
passport bio-data page illuminated from a first direction to show a
first reflection feature;
[0013] FIG. 3B is a schematic illustration of an image of a
passport bio-data page illuminated from a second direction to show
a second reflection feature;
[0014] FIG. 3C is a schematic illustration of an image of the
passport bio-data page of FIGS. 3A and 3B with no reflection
features visible;
[0015] FIG. 4 is a chart illustrating the pixel intensity of a raw
pixel data set I.sub.PR against distance d from the source of
illumination;
[0016] FIG. 5 is a chart showing the final pixel intensity I.sub.PF
of the pixels in the first set of pixel image data (as an example)
against distance d from the source of illumination;
[0017] FIG. 6 is a chart showing pixel intensity I.sub.P against
apparent greyness G (the response of the image capture device
across the spectrum imaged) for decreasing pixel intensity;
[0018] FIG. 7 is a schematic example of the effect that gamma
correction has on text within an image;
[0019] FIG. 8 is a schematic illustration of a portion of the color
sensor array for an image capture device; and
[0020] FIG. 9 is a flow chart illustrating the preferred embodiment
of the present invention.
DETAILED DESCRIPTION
[0021] Security documents such as passports, identification cards,
and the like, may often have either a matte or a shiny finish, and
is unlikely to be completely flat. During use, corners of plastic
bio-data pages in passports, for example, may bend, air bubbles and
dirt may become trapped within a laminate structure, or a surface
material may be highly reflective and shiny in appearance, all of
which can create unwanted reflections, generally specular
reflections, thus distorting the captured image. This may make
machine readable text, such as OCR text, overt and covert security
features difficult to read, and make automatic authentication of
the document and/or verification of the holder unreliable or
impossible. For example, a bio-data page having a laminate
construction with an extremely shiny surface may require additional
inspection by an operator if specular reflections distort the image
beyond the capability of an automatic reader.
[0022] With the various constraints on security document imaging in
mind, there is a need for a method that allows the image taken by a
standard security document reader to be enhanced sufficiently that
stray and unwanted reflections are no longer an issue, such that
the document can be authenticated reliably and accurately
regardless of surface quality or illumination conditions. Such a
method may also find applications in other image capture
techniques.
[0023] The present invention aims to address these issues by
providing an image enhancement method for an image capture device;
the method comprising the steps of: illuminating an object placed
on, in or adjacent to the image capture device and capturing an
image of the object from a first position to obtain a first set of
raw pixel data; illuminating the object placed on, in or adjacent
to the image capture device, capturing an image of the object from
a second position, different to the first position, to obtain a
second set of raw pixel data, where each pixel in the second set of
raw pixel data corresponds to a pixel in the first set of raw pixel
data representing a point on the object; calibrating each of the
first and second sets of raw pixel data using a first set of image
calibration pixel data to create a first set of image pixel data
and using a second set of image calibration pixel data to create a
second set of pixel image data respectively, where each pixel in
the second set of image calibration pixel data corresponds to a
pixel in the first set of image calibration pixel data, and each
pixel in the first and second sets of image calibration pixel data
corresponds to a pixel in each of the first and second sets of raw
pixel data respectively; for a first pixel in the first set of
image pixel data: i) calculating the modulus of the pixel intensity
of said pixel in the first set of image pixel data minus the pixel
intensity of the corresponding pixel in the second set of image
pixel data and comparing the modulus to a pre-determined threshold
value; ii) if the modulus is greater than the threshold value,
selecting the one of the pixel in the first set of image pixel data
and the pixel in the second set of image pixel data with the lowest
pixel intensity as the output pixel; iii) if the modulus is less
than the threshold value, determining the one of the corresponding
pixel in the first set of image calibration pixel data and the
corresponding pixel in the second set of image calibration pixel
data having the highest pixel intensity, and selecting the
corresponding pixel in either the first set of image pixel data or
the second set of image pixel data as the output pixel; and
repeating steps i), ii) and iii) for at least a second pixel in the
first set of image pixel data and forming a set of final pixel data
from the resulting output pixels.
[0024] The advantage of using such an approach is that only pixels
representing a portion of an image in which a reflection is absent
are used to make up the set of final image data, thus ensuring that
any image recovered is of a high quality with reflections either
attenuated or removed. In some circumstances, reflections may in
fact be separated, for example, specular reflections are removed
but reflections from single color features remain. This is
particularly advantageous for a security document, such as an
identity document or a fiduciary document, where covert or overt
security features may be revealed as single color reflections.
[0025] In one aspect of the present invention, the object is
illuminated with visible light, infra-red light or ultraviolet
light. In another aspect of the present invention, when the object
is illuminated with visible light, the object is illuminated with
white light.
[0026] In another aspect of the present invention, the pixel
intensity may have balanced red-green-blue components.
Alternatively, the pixel intensity has un-balanced red-green-blue
components. In this case, preferably the pixel intensity has a
maximum red, green or blue component.
[0027] In another aspect of the present invention, the method may
further comprise: for each pixel in the first and second sets of
raw pixel data, measuring the intensity of single color
reflections, and for pixels representing the same region of the
object, selecting the pixel with the brightest single color
intensity; and including said pixel in a second set of final image
data.
[0028] In yet another aspect of the present invention, the method
may also further comprise the step of adjusting the first and
second sets of raw image data with a gamma correction.
[0029] In another aspect of the present invention, the image
enhancement output includes the attenuation, separation or removal
of reflections. In yet another aspect of the present invention, the
image enhancement output includes the attenuation, separation or
removal of specular reflections. In another aspect of the present
invention, the method may also further comprise: for each of the
first and second sets of raw pixel data, compensating the intensity
values of each pixel for ambient light. In this case, the method
may also further comprise: creating a set of ambient pixel data by
imaging the object under no illumination other than ambient light;
and subtracting the set of ambient pixel data from each of the
first and second sets of raw pixel data.
[0030] Preferably the object is a security document. More
preferably the object is an identity document or a fiduciary
document. Yet more preferably the object is one of a passport, an
identification card or a driver's licence. In each of these
situations, preferably the image capture device is a security
document reader.
[0031] In the present invention, the use of masking techniques are
used to create an image of an object, such as a security document,
that is substantially free of unwanted reflections, as explained in
further detail below. The method comprises illuminating an object
placed on, in or adjacent to an image capture device such as a
security document reader, so as to capture an image of the object
from a first position to obtain a first set of raw pixel data.
Next, a step of illuminating the object placed on, in or adjacent
to the image capture device, capturing an image of the object from
a second position is carried out, where the second position is
different to the first. This is to obtain a second set of raw pixel
data, where each pixel in the second set of raw pixel data
corresponds to a pixel in the first set of raw pixel data
representing a point on the object. Since there is a positional
difference in the two images, calibrating each of the first and
second sets of raw pixel data is preferred. This is done using a
first set of image calibration pixel data to create a first set of
image pixel data and using a second set of image calibration pixel
data to create a second set of pixel image data respectively, where
each pixel in the second set of image calibration pixel data
corresponds to a pixel in the first set of image calibration pixel
data, and each pixel in the first and second sets of image
calibration pixel data corresponds to a pixel in each of the first
and second sets of raw pixel data respectively. At this point, it
is preferred to carry out an operation to mask the reflections
detected to produce an enhanced image of the object. Starting with
a first pixel in the first set of image pixel data, the modulus of
the pixel intensity of the pixel in the first set of image pixel
data minus the pixel intensity of the corresponding pixel in the
second set of image pixel data is calculated, and compared to the
modulus to a pre-determined threshold value. If the modulus is
greater than the threshold value, the one of the pixel in the first
set of image pixel data and the pixel in the second set of image
pixel data with the lowest pixel intensity is selected as the
output pixel. If the modulus is less than the threshold value, then
the one of the corresponding pixel in the first set of image
calibration pixel data and the corresponding pixel in the second
set of image calibration pixel data having the highest pixel
intensity is determined. The corresponding pixel in either the
first set of image pixel data or the second set of image pixel data
is selected as the output pixel. These steps are then repeated for
sufficient pixels in the first set of image pixel data, so as to
render an image from a set of final pixel image data formed from
the resulting output pixels.
[0032] By using such a masking technique described above, an image
substantially without reflections is revealed. Such a method is
particularly suitable for use with a security document. By
attenuating, separating or removing unwanted reflections, in
particular, specular reflections, the reliability of automated
authentication of a security document, either by text or overt
security feature recognition or by revelation of covert security
features is improved.
[0033] In the following embodiments, the example of a security
document and security document reader is used. However, as
described below, in alternative embodiments, the method of the
present invention suitable for use with other objects and image
capture devices.
[0034] FIG. 1 is a schematic side view of a document reader in
which an embodiment of the method of the present invention is
carried out. The document reader 1 is generally cuboid in shape,
and comprises a housing 2 in which first 3 and second 4
illumination sources and an image capture device 5 are positioned.
The uppermost surface of the housing 2 is formed from a glass
platen 6, onto which a security document 7 may be placed in order
to be imaged. In this embodiment, in order to enable illumination
of the document from a first and a second direction, the first 3
and second 4 lighting sources are positioned on either side of the
image capture device 5, which is disposed centrally within the
housing 2 adjacent a wall 8 of the housing. Each illumination
source 3, 4 is provided with a linear array of light emitting
diodes 9a, 9b, 9c, 9d (only two of which are shown on each of the
first 3 and second 4 lighting sources for clarity), aligned to
illuminate the entire surface of a security document 7 in contact
with the glass platen 6. Light travels along the optical paths
OP.sub.1 and OP.sub.2 to be incident on the glass platen 6 and
document 7, and reflected back to the surface of the image capture
means 5. Non-limiting example optical paths are shown for the first
illumination means 3 only. Second illumination source 4 may include
similar optical paths, although not illustrated. Preferably the
light emitting diodes emit light in the visible range of the
electromagnetic spectrum, with suitable LEDs being available from
Osram Opto Semiconductors under the product code "TOPLED Ultra
White 2PLCC". The image capture means 5 is preferably a CMOS
device, such as the MT9T001 1/2 inch 3-megapixel digital image
sensor, available from Micron Technologies, Inc., located in Boise,
Id., USA.
[0035] The document reader 1 illustrated in FIG. 1 is arranged so
as to enable a method involving imaging a security document from a
first and a second direction, where the second direction is
different from the first direction. Using two different
illumination directions allows images of the same point on the
security document to be taken that yield different optical effects.
This is generally illustrated in FIG. 2. FIG. 2 is a schematic side
view of an optical defect in a security document giving rise to a
specular reflection. Specular reflections may be anything that
includes an optical glare reflecting back from a surface. Examples
of specular reflection in a security document may be caused an
uneven laminate, uneven surface that is not optically flat, or the
material itself, such extra shiny laminates. In general, specular
reflections are mirror or glass-like reflections. In the case of a
security reader, there are artifacts or material properties in the
security laminates of a security document that cause bright white
spots where the light from the light source(s) is reflected back to
the image capture device. As one example, optical defect 10 is
present in the surface of a security document 11, in this case, a
bubble in a laminated bio-data page structure. Light from a first
direction L.sub.1 is incident on a first side of the defect 10 and
reflected R.sub.1 onto an image capture device 12. This gives an
image with a bright spot corresponding to reflection from the
surface of the defect 10 on which the light L.sub.1 was incident.
Light from a second direction L.sub.2 is incident on a first side
of the defect 10 and reflected R.sub.2 onto an image capture device
12. This gives an image with a bright spot corresponding to
reflection from the surface of the defect 10 on which the light
L.sub.2 was incident. These two images of the same section of the
security document 11 will appear to be subtly different when
compared to each other. When light reflected from the defect 10 is
incident on the image capture device 12 different pixel intensities
for the same point on the security document are obtained as
follows. The image capture device 12 contains an array of cells
each of which has a one-to-one relationship with a pixel in an
image of the security document 11. When illuminated from a first
direction with light L.sub.1 the first set of raw pixel data
obtained will contain a bright pixel at the point where the
reflection R.sub.1 is incident on the image capture device, at
position A. When illuminated from a second direction with light
L.sub.2 the second set of raw pixel data obtained will contain a
bright pixel at the point where the reflection R.sub.2 is incident
on the image capture device, at position B. When these two data
sets are combined the darkest pixel (e.g. the pixel with the lowest
pixel intensity measure for each equivalent pixel) will be found in
the second raw pixel data set at point A and in the first raw pixel
data set at point B. A final image formed from combining data based
on these two data sets and using only the "darkest" pixels or
pixels with the lowest pixel intensity measured at each position on
the security document such that reflections from the defect 10 are
effectively removed from this final image. This is possible as each
pixel in the second set of raw pixel data corresponds to a pixel in
the first set of raw pixel data representing a point on the
security document.
[0036] This idea is illustrated further in FIGS. 3A, 3B and 3C.
FIG. 3A is a schematic illustration of an image of a passport
bio-data page illuminated from a first direction L.sub.1 to show a
first reflection feature. A bio-data page is chosen in this example
as typically this is comprised of a multilayer laminated structure
with at least one plastic or reflective layer or region on the page
containing identity information about the passport bearer. However,
the method described below is equally suitable for any page or
surface of a security document that requires imaging for bearer
identification and/or document authentication to take place. The
first reflection feature 13 is a specular reflection obscuring a
portion of text 14 on the bio-data page 15. This is caused, for
example, by a defect within the laminated structure of the bio-data
page 15. FIG. 3B is a schematic illustration of an image of a
passport bio-data page illuminated from a second direction L.sub.2
to show a second reflection feature. The second reflection feature
16 is a specular reflection obscuring a portion of the photograph
17 of the holder of the bio-data page 15. This is caused, for
example, by the inclusion of a reflective covert security feature
within the bio-data page 15. FIG. 3C is a schematic illustration of
an image of the passport bio-data page of FIGS. 3A and 3B with no
reflection features visible. This image is formed from a comparison
of the two images in FIGS. 3A and 3B and using a masking technique
to select pixels revealing a reflection free image.
[0037] In order to utilise a masking approach to its fullest
extent, it is preferably to ensure that the data collected in the
first and second raw pixel data sets is as accurate as possible. To
achieve this, two factors must be born in mind. Firstly, a document
reader, such as a security document reader, has a limited footprint
due to size restrictions in the environment in which it is used,
which would typically be a desk or cubicle at a border inspection
point. This then places constraints on the optical system within
the document reader, as to enable illumination of an entire
security document placed on the reader lighting source often need
to be positioned adjacent a wall or corner of the housing of the
document reader, as in the example given above. This causes a
variation in the intensity of illumination of the security document
with distance away from the lighting means, and consequently a
spatial distribution of pixel intensity in image obtained.
Secondly, the image capture device typically has an inherent
non-linear response to intensity of illumination and color, leading
to a variation between a real intensity for a particular shade and
an ideal intensity for the same shade. For a methodology that
relies on being able to select the darkest version of a pixel any
discrepancy in illumination and/or color definition can have a
detrimental effect on the data unless corrected.
[0038] FIG. 4 is a chart illustrating the pixel intensity of a raw
pixel data set I.sub.PR against distance from the source of
illumination d. This illustrates the effect of the spatial
distribution of the light emitted from the lighting sources 3, 4,
within the document reader 1 and incident on the security document
7. In this example, the lighting source 3 is positioned adjacent
d=0, such that the highest pixel intensity of raw pixel data
I.sub.PR occurs at this point. As the distance d away from the
lighting source 3 increases the pixel intensity drops off
substantially following a mean inverse square approximation. The
relationship shown is appropriate for two lighting sources, whereas
for a greater number of lighting sources the resulting intensity
relationship is created using a mean inverse square approach
resulting in a saddle-shaped intensity distribution. In this
example, approximately half-way between the highest and lowest
pixel intensities a reflection peak RP is seen. However, given the
general noise within the data and the decreasing pixel intensity
with distance d in this position it is likely that the reflection
peak would be detected and the darkest pixel method used
successfully. However, a peak found at an increased value of d, and
therefore further into the region of decreasing pixel intensity may
be harder to detect due to noise, and therefore calibration of the
raw pixel data to avoid this is advisable.
[0039] This variation in pixel intensity can be corrected using a
set of calibration pixel data. Each of the first and second raw
pixel data sets will have an intensity distribution similar to that
shown in FIG. 4. Also shown on FIG. 4 is a line marked "WBG"
representing white background intensity. This is effectively the
pixel intensity for a plain white background, such as a sheet of
white paper or card, imaged using the document reader 1. By
allocating pixel intensity in the WBG to represent the background
value of pixel intensity in the raw pixel data sets, a set of
calibration pixel data is created. When this set of calibration
pixel data is combined with the raw pixel data in a mathematical
operation as shown in Equation 1 below the pixel image data is
returned:
Output=(255.times.Input)/(WBG+c) Equation 1
Output=output pixel intensity in pixel image data Input=input pixel
intensity in raw pixel data 255=maximum intensity value allocated
to the cell in the image capture device WBG=intensity of
corresponding pixel in calibration pixel data c=constant, greater
than 0 and preferably 1, included to ensure that the Output value
is not infinite.
[0040] This operation is completed for both the first set of raw
pixel data and the second set of raw pixel data to obtain the first
and second sets of image pixel data respectively. FIG. 5 is a plot
showing the final pixel intensity I.sub.PF of the pixels in the
first set of pixel image data (as an example) against distance from
the source of illumination d. It can be seen that the background
intensity is now substantially flat with increasing distance, and
the reflection peak RP seen clearly above the background intensity,
allowing the darkest pixel to be chosen easily and accurately.
Since the two sets of raw pixel data are different, calibrating
each of the first and second sets of raw pixel data comprises using
a first set of image calibration pixel data to create a first set
of image pixel data and using a second set of image calibration
pixel data to create a second set of pixel image data respectively.
Each pixel in the second set of image calibration pixel data
corresponds to a pixel in the first set of image calibration pixel
data, and each pixel in the first and second sets of image
calibration pixel data corresponds to a pixel in each of the first
and second sets of raw pixel data respectively.
[0041] The calibrated first and second sets of image pixel data may
then be used to calculate a first set of final image data by using
a masking technique to select between corresponding pixels in an
image of the same point on the image of the security document. The
masking technique is intended to remove the effects of "background
noise" to successfully identify pixels representing reflections
within a particular set of image pixel data. To enable this, a
thresholding operation is carried out, where an arbitrarily chosen
threshold is used to remove any noise and to identify bright
pixels, thus effectively creating a mask. By considering the
modulus of the difference in intensity between a first pixel in the
first set of image pixel data and the corresponding pixel in the
second set of image pixel data, it will be obvious if a reflection
is present at such a position on the security document since the
difference in intensity between the two pixels will be high. If
this is greater than the threshold (more than background noise) it
can be assumed that the brighter pixel represents a reflection,
hence to retrieve an image where no reflection is present the
"darkest" or lowest intensity pixel of the two is used as the
output pixel, and included in the set of final image pixel data.
However, if the modulus is below the threshold, a reflection is
absent, and hence another criterion must be chosen to determine
which of the pixels in the first image pixel data set or the second
image pixel data set should be used in the set of final image
data.
[0042] In an embodiment of the present invention, this is done by
assessing the value of the corresponding pixels in the first and
second calibration pixel sets. The calibration pixel data sets
represent the background intensity rather than any artefact of the
security document being imaged. In addition, there is greater noise
in the pixels in a particular raw pixel data set with increasing
distance d from the lighting means. This has the effect that, when
a pixel is chosen, merely choosing the darkest pixel direct from
the first image pixel data set or the second image pixel data set
may lead to a poor quality pixel being chosen that yields little
benefit in terms of final image quality. However, for regions where
there is no reflection present, by determining which pixel in the
first calibration data set and the second calibration set has the
highest intensity and selecting the appropriate corresponding pixel
from either the first or second image pixel data sets, the
brightest pixel giving the best quality final image possible is
yielded.
[0043] The operation may be done using a simple code loop as
follows:
TABLE-US-00001 If |P.sub.1 - P.sub.2| > Threshold then if
P.sub.1 > P.sub.2 P.sub.output = P.sub.2 else P.sub.output =
P.sub.1 else if W.sub.1 > W.sub.2 P.sub.output = P.sub.1 Else
P.sub.output = P.sub.2
Where P.sub.1 is a pixel in the first set of image pixel data,
P.sub.2 is the corresponding pixel in the second set of image pixel
data. P.sub.output is the output pixel forming part of the final
set of image pixel data. W.sub.1 is the pixel in the first set of
calibration data corresponding to the pixel in the first set of
image pixel data, and W.sub.2 is the pixel in the second set of
calibration data corresponding to the pixel in the second set of
image pixel data. W.sub.1 and W.sub.2 are therefore also
corresponding pixels. The threshold is chosen arbitrarily, based on
the maximum intensity of the image capture device used. For
example, a CMOS device will typically have a maximum intensity of
255, hence a suitable threshold to remove any background noise
would be approximately 10% of such a maximum, so around 30.
[0044] The code loop is repeated for as many times as is necessary
to form a viable and useful image, which may be for all of the
pixels in the first image pixel data set or for only a subset of
these pixels. The resulting image may be used in various
authentication and verification processes, since any reflections in
the original images are attenuated, reduced or separated to the
extent that further processing operations are reliable and
reproducible.
[0045] However, as is evident from FIG. 6, it may be desirable to
make a further correction, such as a gamma correction, to take into
account the inherent non-linear response to intensity of
illumination and color, leading to a variation between a real
intensity for a particular shade and an ideal intensity for the
same shade. FIG. 6 is a chart showing pixel intensity I.sub.P
against apparent greyness G (the response of the image capture
device across the spectrum imaged) for decreasing pixel intensity.
In the centre of the response range the non-linear behaviour of the
image capture device is at its most stark--with the greatest
deviation being either above (I.sub.1) or below (I.sub.2) the ideal
intensity I.sub.IDEAL. The direction in which the deviation occurs
is an artefact of the image capture device used, hence both upper
and lower deviations are illustrated here for the purposes of
explanation. In order to ensure that the pixel intensity is as
close to the ideal intensity as possible a correction factor, often
known as gamma correction, is used. When applied to the pixel
intensity at point A on curve I.sub.2, the pixel intensity will be
corrected to point A' on the line I.sub.IDEAL, and when applied to
the pixel intensity at point B on curve I.sub.2, the pixel
intensity will be corrected to point B' on the line I.sub.IDEAL.
Gamma correction is an exponential function typically in the form
shown in Equation 2 below:
V.sub.out=AV.sub.in.sup..gamma. Equation 2
[0046] Where V.sub.out is output, V.sub.in is input, A is a
constant and .gamma. is the gamma exponential correction factor. A
gamma correction is applied to the first set of final image data if
required to ensure that the data quality in the first set of final
image data is as high as possible, making it ideal as a starting
point for further processing as part of a document authentication
process. FIG. 7 is a schematic example of the effect that gamma
correction has on text within an image. The upper line of text
contains a first group of letters 18 (all letter "A") corresponding
to low illumination intensity (i.e. at a large distance d from the
lighting source) and thus appear all in a lighter shade of gray,
and a second group of letters 19 (all letter "A") corresponding to
high illumination intensity (i.e. at a small distance d from the
lighting means) and thus appear all in a darker shade of gray. Both
groups 18, 19 are without gamma correction. The lower line of text
contains a third group of letters 20 (all letter "A") corresponding
to low illumination intensity (i.e. at a large distance d from the
lighting source) and a fourth group of letters 21 (all letter "A")
corresponding to high illumination intensity (i.e. at a small
distance d from the lighting source), both with gamma correction.
The effect of gamma correction on an image is that, for the letters
in the third group 20 and fourth group 21, there is a greater
contrast between individual shades, and a greater contrast between
lighter shades (low illumination) and darker shades (bright
illumination) in general (i.e. the contrast between the entire
third group 20 and the entire fourth group 21).
[0047] Extraction of further image features, such as covert
security features hidden within the security document being imaged,
or further correction and enhancement of the raw image pixel data
will now be described with respect to further embodiments of the
present invention.
[0048] Although in the above embodiment no distinction is made in
relation to the color of reflection under examination in a first
further embodiment features may be separated, attenuated,
highlighted or removed by considering brightest single color
intensities as a complement to the masking approach outlined above.
For specular reflections RGB (red, green and blue intensity) values
are typically balanced out creating a bright white spot. However,
for security features, often only one of the RGB values is
maximised, since the feature is brighter in a single color only. So
for the darkest only pixel approach outlined above, the pixel
intensity has balanced red-green-blue components, since this
corresponds to a white, specular reflection. For a security
feature, the pixel intensity has un-balanced red-green-blue
components. This may in fact be that pixel intensity has a maximum
red, green or blue component.
[0049] FIG. 8 is a schematic illustration of a portion of the color
sensor array for an image capture device. This is typical of a
CMOS-type device used in the embodiment above. Sensors are grouped
into groups of four each comprising a red detector cell
(R.sub.1-R.sub.8), a blue detector cell (B.sub.1-B.sub.8) and two
green detector cells (G.sub.1-G.sub.8, G'.sub.1-G'.sub.8),
representing a cell having a one-to-one relationship with a pixel
in a final image. Each sensor detects the appropriate color, with
two green detector sensors being included in each group to mimic
the response of a human eye. The color response of a reflection,
i.e. determination of a pixel having the brightest single color
intensity, is measured by considering the response of individual
sensors within each group and adjacent sensors within each group
and/or adjacent groups. For example, a reflection with an intense
blue component can be detected by merely looking at the response of
the blue detector sensors or the red and green detector sensors
(for the presence or absence of a response) or by looking at the
response of adjacent blue detector sensors. For example, saturation
of the blue B.sub.2 sensor would result in the response of the blue
B.sub.1, B.sub.3 and B.sub.5 sensors being examined as strong
response here would indicate a reflection peak. Consequently, by
additionally measuring the color intensity of single color
reflections by examining the color response of the pixels in the
first and second sets of raw pixel data, for pixels representing
the same point on the security document, the pixel with the
brightest single color intensity can be selected and included in
the first set of final image data. As an alternative to using the
RGB color space it may be desirable to use a different color space,
such as L*a*b*, since this mimics the natural response of the eye
more accurately than RGB space, which is advantageous when an
operator compares images on a screen and the actual security
document.
[0050] In the examples given above, no correction is required for
the effects of ambient lighting (i.e. light generated by the
surrounds of the document reader rather than by the document
reader), since typically document readers are used in an enclosed
situation, for example, by providing a hood or lid covering the
security document during illumination. However, in some
circumstances, such as when a document reader is used in a booth or
other open environment, it may be desirable to correct the image
obtained by removing the intensity component attributable to
ambient light. In a further embodiment of the present invention,
this is done by creating a set of ambient pixel data by imaging the
security document under no illumination other than ambient light.
This may be achieved by placing the security document onto the
glass platen 6 of the document reader 1 and without activating any
of the lighting sources, capturing an image of the security
document 7, thus creating the set of ambient pixel data. This set
of ambient pixel data is then subtracted from each of the first and
second sets of raw pixel data. This may be done at the same time as
other calibration operations, beforehand or afterwards, but before
the first or second sets of final image data are created.
[0051] As an example, the reflection removal technique was carried
out using a commercially available security document reader, a
QS1000 available in the UK from 3M United kingdom PLC, 3M Centre,
Cain Road, Bracknell, Berkshire, RG12 8HT, UK. Minor modifications
were made to the reader to split the existing array of
light-emitting diodes (LEDs) into two separate half-arrays to
ensure that two separate lighting sources were created. This was
done by physically re-wiring the circuit board and including
additional code in the software controlling the illumination to
allow each half-array to be operated separately. In order to ensure
that there was a one-to-one identity between corresponding pixels
in any data sets obtained using either half-array, a mapping system
was used to uniquely identify pixels. Each pixel was allocated a
unique identifier based on its position with respect to an
arbitrary x-axis corresponding to the front edge of the reader and
an arbitrary-axis corresponding to a side edge of the reader, each
identifier being of the format (x.sub.n, y.sub.n).
[0052] To test the reflection removal technique, the following
steps were carried out, as shown in FIG. 9, a flow chart
illustrating the preferred embodiment of the present invention. At
step 101, a passport was opened to reveal the bio-data page, and
placed face-down on the glass platen of the document reader. At
step 102, the bio-data page was illuminated using the first
half-array of LEDs from a first direction to capture the first raw
pixel data set. At step 103, the bio-data page was illuminated from
a second direction, different to the first, using the second
half-array of LEDs to capture the second raw pixel data set. At
step 104, the first and second sets of raw pixel data were
calibrated using a set of image calibration pixel data to create a
first set of image pixel data and a second set of pixel image data
respectively. The set of calibration data was obtained initially
when the document reader was set up to illuminate from two
different directions by imaging a sheet of white 80 gsm paper. At
step 105, a first set of final image data was calculated by
comparing the first and second sets of image pixel data and for
pixels representing the same point on the object, selecting the
pixel with the lowest pixel intensity; and including said pixel in
the first set of final image data. This was done by using the
following loop:
TABLE-US-00002 For (x.sub.n, y.sub.n) If |P.sub.1 - P.sub.2| >
Threshold then if P.sub.1 > P.sub.2 P.sub.output = P.sub.2 else
P.sub.output = P.sub.1 else if W.sub.1 > W.sub.2 P.sub.output =
P.sub.1 Else P.sub.output = P.sub.2
Repeat for all x (x.sub.1-x.sub.n) and y, (y.sub.1-y.sub.n) to
create the first set of final image data comprising the
OutputPixels for each (x, y). Once the first set of final image
data was obtained, it was necessary to perform a gamma correction
exercise to ensure that any effects of the response of the image
capture device within the reader were minimised. To do this, before
initial use, the image capture device was calibrated using a set of
color reference targets available from X-Rite, 4300 44th St. SE,
Grand Rapids Mich. 49512, USA. The color reference targets comprise
a set of greyscale targets with known RGB values, which in
conjunction with image calibration software allow a matrix of
.gamma. values to be calculated at certain points in the response
of the image capture device. This matrix of .gamma. values was then
applied to the first set of final image data to correct for any
inherent response behaviour in the image capture device.
[0053] Although the technique was carried out using a passport
bio-data page, it is also possible to image any other page of a
passport, an identification card or a driver's licence, as examples
of identity documents. Other security documents, such as fiduciary
documents (for example, credit or bank cards) may also be imaged
using this technique. In the above example, the processing to
create the various sets of data is carried out within the FPGA
(field-programmable gate array) of the document reader. However,
this is merely a matter of preference, and the processing could
alternatively be carried out in an ASIC (application-specific
integrated circuit) if desired.
[0054] In the above embodiments, images are captured from different
positions, such as from different angles. This is dictated by the
physical construction of a passport reader, which has a dedicated
footprint limited in size due to the constraints of the areas in
which such readers are often situated. A typical full page passport
reader has an approximate base size of 160 mm.times.200 mm and a
height of approximately 190 mm. The lighting sources are typically
placed adjacent a side wall, approximately 50 to 70 mm away from
the wall, resulting in a typical angle of illumination in the range
of 10.degree. to 60.degree. and typically around 40.degree. to
50.degree. (where the angle is measured at the surface of the
security document being illuminated). This is relatively wide angle
illumination compared with other image capture devices, such as
cameras. Consequently the first and second positions from which the
security document is illuminated and the images captured from are
determined by the first and second illumination angle created by
the position of the first and second lighting sources. However, it
is possible to create illumination and/or image capture from
different relative positions without using two separate lighting
sources. For example, a single image capture device can be replaced
with two or more image capture devices, in conjunction with a
single lighting source. Alternatively, further optical paths can be
created from either a single or multiple light source(s) using
lenses, mirrors or prisms, with each optical path yielding a
relative position from which the security document may be
illuminated or an image captured. Creating different relative
positions from which to illuminate the security document or from
which to capture images of the security document may also be
achieved by moving the security document and/or the image capture
device relative to each other. This could be using a motor or
vibrating either the security document (for example, by moving the
glass platen) or the image capture device at a fixed frequency.
Creating multiple relative positions from which either the security
document can be illuminated or from which images can be captured is
particularly useful for identifying holographic features. Further
options could also include the use of plenoptic light field cameras
or the use of microlens arrays to create multiple images that
appear to be imaged from multiple angles.
[0055] In the embodiments described above, the approach of the
present invention is applied to a security document reader to
address issues involving reflections in security documents.
However, the techniques may be used with other image capture
devices (including, but not limited to, cameras--whether digital,
video or otherwise--CMOS and CCD devices, mobile phones and other
hand held devices, optical scanners, including flat bed scanners
and other equipment that is capable of capturing an image) in which
reflections arising from optical or physical defects or
inconsistencies in the object being imaged occur. In the
embodiments described above, the security document may be replaced
by an object, for example a different type of document (in the case
of a scanner), or a person or landscape scene (in the case of a
camera). This may or may not be in contact with the image capture
device, and the angle of illumination may be relatively narrow
compared with the example of a passport reader above. However,
illumination of the object or capture of an image of the object
from at least two positions enables the darkest only pixel
technique to be applied to remove reflections in images of the
object. The code loops described above also apply equally well to
other object types and image capture devices, since images of an
object from different positions will always yield at least one
image in which a reflection is present at a certain point and at
least a second image where a reflection is absent at the same
point, hence there will always be one bright and one dark
corresponding pixel.
[0056] The present invention has now been described with reference
to several embodiments thereof. The foregoing detailed description
and examples have been given for clarity of understanding only. No
unnecessary limitations are to be understood therefrom. All patents
and patent applications cited herein are hereby incorporated by
reference. It will be apparent to those skilled in the art that
many changes can be made in the embodiments described without
departing from the scope of the invention. Thus, the scope of the
present invention should not be limited to the exact details and
structures described herein, but rather by the structures described
by the language of the claims, and the equivalents of those
structures.
* * * * *