US6741756B1
Expired / Lapsed
US
⏱️ 0.0 years remaining
System and method for estimating the orientation of an object
Assignee: Microsoft Technology Licensing LLC
•
Filing: 1999-09-30
•
Granted: 2004-05-25
•
Expires: 2019-09-30
Abstract
Abstract The present invention is embodied in a system and method for automatically estimating the orientation or pose of an object, such as a human head, from any viewpoint and includes training and pose estimation modules. The training module uses known head poses for generating observations of the different types of head poses and the pose estimation module receives actual head poses of a subject and uses the training observations to estimate the actual head pose. Namely, the training module receives training data and extracts unique features of the data, projects the features onto corresponding points of a model and determines a probability density function estimation for each model point to produce a trained model. The pose estimation module receives the trained model and an input object and extracts unique input features of the input object, projects the input features onto points of the trained model and determines an orientation of the input object that most likely generates the features extracted from input object.
Inventors
- 👤 Kentaro Toyama
- 👤 Ying Wu
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-09 22:22
CPC Classifications
G; G06; G06V; G06V40/00; G06V40/10; G06V40/16; G06V40/161; G06T; G06T7/00; G06T7/70; G06T7/73; G06T7/75
Claims Analysis
Total Claims: 44
Claim 1
Independent Claim
100 words
A method for estimating the orientation of an object comprising: providing a trained model of the object as training data, wherein the trained model of the object is produced by extracting unique features of the training data, projecting the features of the training data onto corresponding points of a model and determining a probability density function estimation for each model point; extracting unique features of an image representing the object; back projecting the features of the image onto points of the trained model; and determining an estimated orientation of the object that most likely generates the features extracted from object.
Claim 2
Independent Claim
100 words
A method for estimating the orientation of an object comprising: providing a trained model of the object as training data, wherein the trained model of the object is produced by extracting unique features of the training data, projecting the features of the training data onto corresponding points of a model and determining a probability density function estimation for each model point; extracting unique features of an image representing the object; back projecting the features of the image onto points of the trained model; and determining an estimated orientation of the object that most likely generates the features extracted from object.
Claim 3
Dependent Claim
18 words
The method of claim 1 , further comprising using a predictive prior to bias the estimation of orientation.
Claim 4
Dependent Claim
11 words
The method of claim 1 , further comprising preprocessing the image.
Claim 5
Dependent Claim
14 words
The method of claim 3 , wherein preprocessing the image includes resizing the image.
Claim 6
Dependent Claim
20 words
The method of claim 3 , wherein preprocessing the image includes masking the image to filter out unwanted background noise.
Claim 7
Dependent Claim
17 words
The method of claim 3 , wherein preprocessing the image includes performing histogram equalization on the image.
Claim 8
Dependent Claim
24 words
The method of claim 1 wherein extracting unique features includes determining convolution coefficients for each point in the image by applying predefined convolution templates.
Claim 9
Dependent Claim
15 words
The method of claim 1 , wherein extracting unique features includes computing normalized feature vectors.
Claim 10
Dependent Claim
21 words
The method of claim 1 , wherein extracting unique features includes determining dense occurrences of high-frequency components in local image regions.
Claim 11
Dependent Claim
22 words
The method of claim 9 wherein determining dense occurrences of high-frequency components in local image regions is performed with rotationally-invariant Gabor templates.
Claim 12
Dependent Claim
23 words
The method of claim 9 wherein determining dense occurrences of high-frequency components in local image regions with a Gaussian at a coarse scale.
Claim 13
Dependent Claim
24 words
The method of claim 9 , wherein determining dense occurrences of high-frequency components in local image regions a single Gaussian and plural Laplacian templates.
Claim 14
Dependent Claim
16 words
The method of claim 1 , wherein back projecting the features includes rotating the trained model.
Claim 15
Dependent Claim
26 words
The method of claim 13 wherein back projecting the features further includes projecting previously extracted feature vectors of the input image onto the rotated trained model.
Claim 16
Dependent Claim
29 words
The method of claim 14 wherein back projecting the features further includes computing a likelihood that each pose produces observed features using probability density functions of the trained model.
Claim 17
Dependent Claim
32 words
The method of claim 1 , wherein projecting the features onto corresponding points of a predefined model includes using known orientations and rotating the predefined model to coincide with the known orientations.
Claim 18
Dependent Claim
30 words
The method of claim 16 wherein projecting the features onto corresponding points of a model includes projecting previously extracted features of the image onto corresponding points of the rotated model.
Claim 19
Dependent Claim
26 words
The method of claim 17 wherein projecting the features onto corresponding points of a model includes accumulating and storing data for plural images representing the object.
Claim 20
Dependent Claim
14 words
The method of claim 1 wherein the model is a three dimensional ellipsoidal model.
Claim 21
Dependent Claim
79 words
An orientation system for estimating the orientation of an object comprising a pose estimation module that receives a trained model of the object and an input object and extracts unique features of the input object, back projects the features onto points of the trained model and determines an orientation of the input object that most likely generates the features extracted from input object; and wherein the estimation module uses a predictive prior to bias the output estimation of orientation.
Claim 22
Dependent Claim
79 words
An orientation system for estimating the orientation of an object comprising a pose estimation module that receives a trained model of the object and an input object and extracts unique features of the input object, back projects the features onto points of the trained model and determines an orientation of the input object that most likely generates the features extracted from input object; and wherein the estimation module uses a predictive prior to bias the output estimation of orientation.
Claim 23
Dependent Claim
53 words
The orientation system of claim 20 , further comprising a training module that receives training data and extracts unique features of the training data, projects the features of the training data onto corresponding points of a model and determines a probability density function estimation for each model point to produce the trained model.
Claim 24
Dependent Claim
16 words
The orientation system of claim 21 wherein back projecting the features includes rotating the trained model.
Claim 25
Dependent Claim
28 words
The orientation system of claim 22 , wherein back projecting the features further includes projecting previously extracted feature vectors of the input image onto the rotated trained model.
Claim 26
Dependent Claim
31 words
The orientation system of claim 23 , wherein back projecting the features further includes computing a likelihood that each pose produces observed features using probability density functions of the trained model.
Claim 27
Dependent Claim
33 words
The orientation system of claim 21 , wherein projecting the features onto corresponding points of a predefined model includes using known orientations and rotating the predefined model to coincide with the known orientations.
Claim 28
Dependent Claim
32 words
The orientation system of claim 25 , wherein projecting the features onto corresponding points of the model includes projecting previously extracted features of the image onto corresponding points of the rotated model.
Claim 29
Dependent Claim
15 words
The orientation system of claim 20 wherein the model is a three dimensional ellipsoidal model.
Claim 30
Dependent Claim
37 words
The orientation system of claim 20 , further comprising a preprocessing module capable of at least one of resizing the image, masking the image to filter out unwanted background noise and performing histogram equalization on the image.
Claim 31
Dependent Claim
25 words
The orientation system of claim 20 wherein the pose estimation module determines convolution coefficients for each point in the image by applying predefined convolution templates.
Claim 32
Dependent Claim
15 words
The orientation system of claim 20 wherein extracting unique features includes computing normalized feature vectors.
Claim 33
Dependent Claim
22 words
The orientation system of claim 20 , wherein extracting unique features includes determining dense occurrences of high-frequency components in local image regions.
Claim 34
Dependent Claim
24 words
The orientation of system claim 31 , wherein determining dense occurrences of high-frequency components in local image regions is performed with rotationally-invariant Gabor templates.
Claim 35
Independent Claim
78 words
A method for determining the orientation of a head, comprising: extracting unique features of training data representing different heads; projecting the features of the training data onto corresponding points of a model; determining a probability density function estimation for each model point; extracting unique features of the head; back projecting the features of the head onto points of the trained model; and determining an estimated orientation of the head that most likely generates the features extracted from head.
Claim 36
Independent Claim
78 words
A method for determining the orientation of a head, comprising: extracting unique features of training data representing different heads; projecting the features of the training data onto corresponding points of a model; determining a probability density function estimation for each model point; extracting unique features of the head; back projecting the features of the head onto points of the trained model; and determining an estimated orientation of the head that most likely generates the features extracted from head.
Claim 37
Dependent Claim
19 words
The method of claim 33 , further comprising using a predictive prior to bias the output determination of orientation.
Claim 38
Dependent Claim
15 words
The method of claim 33 , wherein the model is a three dimensional ellipsoidal model.
Claim 39
Dependent Claim
22 words
The method of claim 33 , further comprising determining convolution coefficients for each point in the head by applying predefined convolution templates.
Claim 40
Dependent Claim
48 words
The method of claim 33 , wherein back projecting the features includes rotating the trained model, projecting previously extracted feature vectors of the input image onto the rotated trained model and computing a likelihood that each pose produces observed features using probability density functions of the trained model.
Claim 41
Dependent Claim
47 words
The method of claim 33 , wherein projecting the features onto corresponding points of a predefined model includes using known orientations and rotating the predefined model to coincide with the known orientations and projecting previously extracted features of the image onto corresponding points of the rotated model.
Claim 42
Dependent Claim
16 words
The method system of claim 33 , wherein extracting unique features includes computing normalized feature vectors.
Claim 43
Dependent Claim
21 words
The method of claim 33 , wherein extracting unique features includes determining dense occurrences of high-frequency components in local image regions.
Claim 44
Dependent Claim
23 words
The method of system claim 40 wherein determining dense occurrences of high-frequency components in local image regions is performed with rotationally-invariant Gabor templates.
References Cited (Backward Citations)
US5404426A
Cited
US5459636A
Cited
US5850470A
Cited
US6188776B1
Cited
US6044168A
Cited
US6111983A
Cited
US6301370B1
Cited
US6154559A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2007-05-25 | 2007-11-24 | 2008-05-25 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2011-05-25 | 2011-11-24 | 2012-05-25 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2015-05-25 | 2015-11-24 | 2016-05-25 | $7700.00 | Window Closed / Lapsed |