1461048802-3ef15b4b-4303-45b1-bef8-c03f52da4cc6

1. A method for enhancing an image containing endmembers, comprising:
selecting at least one measure of disparity;
partitioning at least one first set of said endmembers in said image into at least
one second set of said endmembers, each said second set incorporating at least
one site having some spectral content,

wherein said at least one site is a generic element of a lattice, and
wherein spatial consistency is imposed on said spectral content of each said site so that each said at least one second set is associated with a spatially smooth region in said image; and
using said at least one measure of disparity, applying a linear mixing model to said at least one second set to globally label said at least one second set,

wherein said method yields improvements in:
assessment of the type and amount of materials in a scene,
unsupervised clustering of a scene, and
post-processing smoothing operations suitable for application to diverse techniques for terrain categorization and classification.
2. The method of claim 1 in which a multi-grid Gibbs-based algorithm is used to partition said at least one first set of said endmembers,
wherein said algorithm is used to estimate an underlying and unobserved process, XP, a discrete labeling Markov Random Field (MRF) process, that associates a label with each said site, and
wherein said XP is used to condition a spectral mixing process (SMP), and
wherein said XP identifies each said site that may be treated as homogeneous, and
wherein said partitioning may proceed with multiple material types incorporated within each said site.
3. The method of claim 2 in which said partitioning using said multi-grid Gibbs-based algorithm proceeds in stages from coarse toward fine grid resolutions,
wherein the number of said stages is defined by the lattice structure of a neighborhood, \u03bes, forming a simple pattern such that a neighborhood of each said lattice element comprises near, intermediate, and far neighbors that are specific multiples of the grid resolution, \u03c3, and
wherein the near neighbors are closest said lattice elements that may touch said site even if only at said site’s corners, and
wherein the next 16 said lattice elements are intermediate neighbors, and
wherein the far neighbors are the 24 said lattice elements on the outside of said next 16 lattice elements, thus forming perimeter layers away from said site of interest at multiples of said \u03c3, and
wherein use of said intermediate and far neighbors facilitates a faster global solution than using just said near neighbors, and
wherein use of said intermediate and far neighbors also enables said algorithm to remember label assignments from prior said stages, and
wherein said \u03c3 determines the spatial sampling of said algorithm at some specific said stage of multi-grid processing, such that at full resolution, i.e., \u03c3=1, said site corresponds to a single said lattice element, and
wherein all coarse resolutions involve neighborhoods of said lattice elements forming a square in which said \u03c3 is the number of pixels along one side of said square.
4. The method of claim 2 in which said Gibbs-based algorithm is initialized conventionally, wherein said initialization enables a post-processing smoothing algorithm that may operate on the output of at least one classification technique.
5. The method of claim 2 in which said partitioning is initialized randomly, wherein said randomly initialized partitioning algorithm is best implemented as a multigrid process using an extended neighborhood system to attain reasonable global labeling accuracy.
6. The method of claim 2 in which said partitioning is initialized using a pre-processing scheme.
7. The method of claim 6 in which said pre-processing scheme is a recognized classification technique.
8. The method of claim 7 in which said recognized classification technique is supervised.
9. The method of claim 7 in which said recognized classification technique is unsupervised.
10. The method of claim 2 in which said multi-grid Gibbs-based algorithm uses an energy function to model distinctions in said image,
wherein, using the Gibbs Equivalent Theorem with an appropriately defined graph, {SP(\u03c3),\u03beP(\u03c3)}, said Gibbs-based algorithm may be derived, and
wherein a maximum a posteriori (MAP) estimate may be computed that maximizes Pr(XP|g) by iteratively sampling from the local Gibbs distribution pertaining to each said site, s\u2208SP(\u03c3), and
wherein computational intensity is reduced and said global labeling is improved.
11. The method of claim 10 in which said MAP estimate of XP, denoted as X*P, is calculated in a first step, termed the Expectation Step, by:
X*P=arg max Pr(XP|X80 ) and
the maximum likelihood estimate (MLE) of a vector of proportions on a label lattice, \u03b2, denoted as {circumflex over (\u03b2)}, is calculated in a second step, termed the Maximization Step, by:
{circumflex over (\u03b2)}=arg max L(\u03b2|X\u03bb,X*P)

wherein said MAP estimate of XP is used as if it were observed, and
wherein complex integration over all possible realizable partitions is avoided by said two-step process.
12. The method of claim 10 in which said MAP estimate is computed using ratios of the relationship:
Pr
\u2062

(
X
s
P

=
x
s
P

\u2758

X
r
P
=

x
r
P
,

r
\u2260
s
)
=

\u2062

Pr
\u2061

(
X
s
P

=
x
s
P

\u2758

X
r
P
=

x
r
P
,

r
\u2208

\u03be
s

P
\u2061

(
\u03c3
)
)
=

\u2062
1

Z
s
\u2062

\u2147
–

1
T
\u2062
U
s

\u2061

(
x
s
P

,

g
_
)
where:
Z
s

=
\u2211
x
s
P

\u2208
\u0393
\u2062

\u2147
–

1
T
\u2062
U
s

\u2061

(
x
s
P

,

g
_
)
and
Us(xsP,g)=energy interaction of site s\u2208SP(\u03c3) with the neighborhood \u03besP(\u03c3).
13. The method of claim 1 in which said at least one measure of disparity is selected from the group consisting essentially of spectral angle, Euclidian distance, classical Kolmogorov-Smirnov measures, mean adjusted Kolmogorov-Smirnov measures, and combinations thereof,
wherein said image is defined with respect to a local neighborhood.
14. The method of claim 13 in which said local neighborhood comprises at least one non-traditional and locally-extended neighborhood.
15. The method of claim 1 in which parameters of said linear mixing model are estimated based on said endmembers in said at least one second set being associated with appropriate said sites therein to enable computation of an enhanced spectral mixing process (SMP),
wherein the set of material spectra, H, is represented as
H={h(1)(\u03bb), . . . , h(Nends)(\u03bb), \u03bb\u2208\u039b}
and Hs\u2282H is the set of material spectra at a pixel site s, NEnds is the total number of said endmembers, and the spectra for the kth said endmember is given by
h(k)=(h(k)(\u03bb1),\u03bb1\u2208\u039b), if h(k)\u2208Hs.
16. The method of claim 1 further employing a simulated annealing method of partitioning hyperspectral imagery, said simulated annealing method initialized by a supervised classification method to provide spatially smooth class labeling for terrain mapping applications.
17. The method of claim 1 providing an estimate of said model as a Gibbs distribution defined over a symmetric spatial neighborhood system that is based on an energy function characterizing spectral disparities in both Euclidean distance and spectral angle, wherein said energy function depends on distance measures termed disparity metrics that provide a measure of the dissimilarity between an individual said site, s, and said individual site’s neighbor, r.
18. An efficient and accurate method for extracting features from hyperspectral data representing a scene by implementing a spectrally-optimal supervised classification algorithm that has been smoothed by a post-processing routine that imposes spectralspatial constraints defined by the Gibbs prior probability distribution, Pr(XP).
19. A method for extracting features from hyperspectral data representing a scene by initializing a partitioning algorithm with the results of a classifier to improve classification by providing initial estimates based on labels that, as applied under the proper conditions, are spectrally optimal, comprising:
a) performing a classification of the scene using a supervised classifier that selects the labeling {circumflex over (x)}sP at each said site s that maximizes the probability Pr(Xs\u03bb|XsP) of a process that is approximately of the form of
Xs\u03bb|XsP=arg max N(\u03bcxsP, \u03a3xsP)

where:
\u03bcxsP=the mean of the soectral vectors for the class label, xsP
\u03a3xsP=the covariance of the specttral vectors; and

b) initializing by the results in Step a), performing a spectralspatial partitioning of said scene using the relationship:
Pr
\u2062

(
X
s
P

=
x
s
P

\u2758

X
r
P
=

x
r
P
,

r
\u2260
s
)
=

\u2062

Pr
\u2061

(
X
s
P

=
x
s
P

\u2758

X
r
P
=

x
r
P
,

r
\u2208

\u03be
s

P
\u2061

(
\u03c3
)
)
=

\u2062
1

Z
s
\u2062

\u2147
–

1
T
\u2062
U
s

\u2061

(
x
s
P

,

g
_
)
where:
Z
s

=
\u2211
x
s
P

\u2208
\u0393
\u2062

\u2147
–

1
T
\u2062
U
s

\u2061

(
x
s
P

,

g
_
)
Us(xsP,g)=energy interaction of site s\u2208SP(\u03c3) with the neighborhood \u03besP(\u03c3).
20. A method employing a simulated annealing method of partitioning hyperspectral imagery for extracting features from hyperspectral data representing a scene, comprising:
using a Bayesian framework to develop a 2-step supervised Gibbs-based classification algorithm,

wherein said algorithm is capable of performing high quality spatially smooth labeling of said hyperspectral data; and
using a linear classifier to initialize said Gibbs-based partitioning algorithm,

wherein said initializing results in improved label accuracy and smoothness as compared to using only said linear classifier without said Gibbs-based partitioning algorithm, and
wherein global labeling accuracy is increased as compared to a stand-alone randomly initialized said Gibbs-based partitioning algorithm, and
wherein said initializing said Gibbs-based partitioning algorithm with said linear classifier also reduces the computation by eliminating the need for a multi-grid process and by allowing said simulated annealing to start at a cooler temperature.

The claims below are in addition to those above.
All refrences to claim(s) which appear below refer to the numbering after this setence.

1. A laryngoscope comprising a handle and an arm;
the arm comprising walls, a coupling area, a light, and a camera, said light and camera being confined within said walls and wherein at least one of said walls is transparent;
said handle comprising first and second ends, and an actuator, the actuator being biased towards a first position and moveable to a second position, said first position being adjacent to said handle first end; and
said actuator further comprising an actuator end, said actuator end extending from said handle second end into said coupling area when the actuator is in the second position
wherein said arm extends from said second end of said handle along a curved path from a proximal end of said arm to a distal end of said arm;
wherein a return spring having a spring end secured to said second end of said handle biases said actuator to said first end of said handle;
and wherein a curved sheath encloses, along said curved path, and cooperatively couples with, said arm in a secure manner when the actuator is in said first position and is uncoupled from said arm when the actuator is moved to said second position.
2. The laryngoscope of claim 1 wherein said arm is comprised of a spring.
3. The laryngoscope of claim 1 wherein said arm is further comprised of a heating unit.
4. The laryngoscope of claim 1, said first and second handle ends comprising a shared straight longitudinal axis, said actuator, when moved between said first and second positions, traveling parallel to said straight longitudinal axis.
5. A laryngoscope comprising a handle, an arm, and a sheath;
the arm comprising walls, a coupling area, a light, and a camera, said light and camera being confined within said walls and wherein at least one of said walls is transparent;
wherein said sheath is slideably and removably coupled to the arm;
said handle further comprising handle first and second ends and an actuator, the actuator being biased towards a first position and moveable to a second position, said first position being adjacent to said handle first end; and
said actuator further comprising an actuator end, said actuator end extending from said handle second end into said coupling area when the actuator is in the second position
wherein said arm extends from said second end of said handle along a curved path from a proximal end of said arm to a distal end of said arm;
wherein a return spring having a spring end secured to said second end of said handle biases said actuator to said first end of said handle;
and wherein said sheath is curved along said curved path, encloses, and cooperatively couples with said arm in a secure manner when the actuator is in said first position and said sheath is uncoupled from said arm when the actuator is moved to said second position.
6. The laryngoscope of claim 5 wherein the arm comprises a sheath connector ridge, wherein said sheath comprises one or more clasps comprising one or more prongs, and wherein said sheath connector ridge and said one or more clasps are structured and arranged to allow cooperative coupling of said sheath with said arm.
7. The laryngoscope of claim 6 wherein said arm is comprised of a spring.
8. The laryngoscope of claim 5 wherein said arm is further comprised of a heating unit.
9. The laryngoscope of claim 5, said first and second handle ends comprising a shared straight longitudinal axis said actuator, when moved between said first and second positions, traveling parallel to said straight longitudinal axis.
10. The laryngoscope of claim 5 wherein said sheath is comprised of a canal capable of being threaded with a bougie.
11. The laryngoscope of claim 10 wherein said canal is open on a longitudinal side.
12. The laryngoscope of claim 5 wherein said sheath is at least partially transparent.
13. The laryngoscope of claim 6 wherein said sheath is comprised of a canal capable of being threaded with a bougie.
14. The laryngoscope of claim 6 wherein said one or more clasps are scored and deformed when the actuator is moved to said second position.
15. The laryngoscope of claim 5 wherein said laryngoscope is capable of being connected to a display unit and a power source, said display unit being comprised of a screen.
16. The laryngoscope of claim 5 wherein said laryngoscope is comprised of motion sensors such that said laryngoscope is capable of being motion activated and powered on upon movement.
17. The laryngoscope of claim 5 wherein said laryngoscope is comprised of motion sensors such that said laryngoscope is capable of being motion activated and powered off when no movement is detected for a predetermined period of time.