Image processing apparatus

• US 7,590,266 B2
• Filed: 09/01/2004
• Issued: 09/15/2009
• Est. Priority Date: 02/18/2004
• Status: Expired due to Fees

• Current Assignee: Fuji Xerox Company Limited (Fujifilm Holdings Corporation)
• Original Assignee: Fuji Xerox Company Limited (Fujifilm Holdings Corporation)
• Inventors: Kato, Noriji, Fukui, Motofumi
• Primary Examiner(s): Bella; Matthew C
• Assistant Examiner(s): Newman; Michael A
• Application Number:
  US10/930,908
• Publication Number:
  US 20050180610A1
• Time in Patent Office: 1,840 Days
• Field of Search: 382/115, 382/117, 382/118, 382/160, 382/215, 382/225
• US Class Current: 382/118
• CPC Class Codes: G06V 10/50 by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
• CPC Class Codes: G06V 40/169 Holistic features and representations, i.e. based on the facial image taken as a whole

Abstract

An image processing apparatus is provided that is capable of executing identification of an object by using a two-dimensional-image having a relatively low resolution. In the image processing apparatus, image data obtained by photo capturing a prescribed object is used as a processing target; with respect to at least one plane relating to a part of the object, an N-dimensional estimated feature value (N ≧3) defining the plane is operated; the N-dimensional estimated feature value and information identifying the original object are associated with each other and stored as a recognition database in a storage unit; and the recognition database is applied to a recognition process of the object.

Claims

1. An image processing apparatus comprising:

means for photo-capturing objects;

a calculating unit; and

a storage unit including a recognition database, wherein

the calculating unit converts pixels of single two-dimensional-image data showing each object into an N-dimensional converted values and calculates N-dimensional estimated feature values which are parameters for expressing a group of the N-dimensional converted values by the following expression,

$\begin{array}{cc}\underset{j=1}{\overset{m}{\sum <br><br>}}<br><br><br><br>{\mathrm{\xi <br><br>}}_j<br><br>p<br><br>\left(x;<br><br>{\mathrm{\theta <br><br>}}_j\right)& \left(1\right)\end{array}$

where;

x denotes the N-dimensional converted value;

ξ

_j and θ

_j are included in the N-dimensional estimated feature values;

p(x;

θ

_j) represents an element distribution;

N is an integer larger than 2; and

m is a natural number,

wherein the calculated N-dimensional estimated feature values for each object maximizes a likelihood that the group of N-dimensional converted values for each object is expressed by the expression (1),

the storage unit stores in the recognition database N-dimensional estimated feature values for each object and object identifying information, which indicates each object, in association with each other,

the calculating unit executes a recognition process that compares N-dimensional estimated feature values for an aimed object with the N-dimensional estimated feature values for the objects stored in the recognition database,

each N-dimensional converted value includes a position coordinate value of the corresponding pixel and a pixel value of the corresponding pixel.

6. The image processing apparatus according to claim 1, wherein each element distribution p(x;

θ

_j) is a probability distribution having a single peak.

8. The image processing apparatus according to claim 1, wherein expression (1) is mixed Gaussian distributions.

2. An image processing apparatus comprising:

a calculating unit that is configured to be accessible to a storage unit including a recognition database holding a plurality of prepared objects; and

an output unit, wherein

the recognition database stores object identifying information for identifying each prepared object and N-dimensional estimated feature values for each prepared object in association with each other, wherein the N-dimensional estimated feature values for each prepared object are parameters for expressing a group of N-dimensional converted value for each prepared object by the following expression,

$\begin{array}{cc}\underset{j=1}{\overset{m}{\sum <br><br>}}<br><br><br><br>{\mathrm{\xi <br><br>}}_j<br><br>p<br><br>\left(x;<br><br>{\mathrm{\theta <br><br>}}_j\right)& \left(1\right)\end{array}$

where;

x denotes the N-dimensional converted value;

ξ

_j and θ

_j are included in the N-dimensional estimated feature values;

p(x;

θ

_j) represents an element distribution;

N is an integer larger than 2; and

M is a natural number,

wherein the N-dimensional converted values for each prepared object include position coordinate values of pixels of single two-dimensional-image data showing each prepared object,

the calculating unit converts pixels of single two-dimensional-image data showing an aimed object for a reorganization into N-dimensional converted values including position coordinate values of pixels of the single two-dimensional-image data showing the aimed object and pixel values of the pixels of the single two-dimensional-image data showing the aimed object,

the calculating unit calculates N-dimensional estimated feature values which are parameters for expressing a group of the N-dimensional converted values for the aimed object by the expression (1),

the calculating unit compares the N-dimensional estimated feature values for the aimed object with the N-dimensional estimated feature values for the prepared objects stored in the recognition database, and

the output unit outputs a result of the comparison.

3. The image processing apparatus according to claim 2, wherein

the single two-dimensional-image data showing the aimed object is obtained by photo capturing the aimed object, and

the calculating unit modifies the N-dimensional estimated feature values for the prepared objects based on a photo captured state of the aimed object.

7. The image processing apparatus according to claim 2, wherein each element distribution p(x;

θ

_j) is a probability distribution having a single peak.

9. The image processing apparatus according to claim 2, wherein the expression (1) is mixed Gaussian distributions.

4. A method of image processing comprising:

using a computer to perform the steps of;

(a) loading single two-dimensional-image data, each showing a prepared object;

(b) converting the two-dimensional-image data into multidimensional variables, each multidimensional variable comprising a positional coordinate value of a corresponding pixel of the two-dimensional-image data and a pixel value of the corresponding pixel of the two-dimensional-image data, the pixels being contained in the single two-dimensional-image data;

(c) calculating multidimensional estimated feature values that are parameters for expressing a group of the multidimensional variables by the following expression

$\begin{array}{cc}\underset{j=1}{\overset{m}{\sum <br><br>}}<br><br><br><br>{\mathrm{\xi <br><br>}}_j<br><br>p<br><br>\left(x;<br><br>{\mathrm{\theta <br><br>}}_j\right)& \left(1\right)\end{array}$

where;

x denotes the multidimensional variable;

ξ

_j and θ

_j are included in the multidimensional estimated feature values;

p(x;

θ

_j) represents an element distribution; and

m is a natural number,

wherein the calculated multidimensional estimated feature values maximize a likelihood that the group of multidimensional variables is expressed by the expression (1);

(d) executing a recognition process by comparing multidimensional estimated feature values for an aimed object with the multidimensional estimated feature values for the prepared objects stored in a recognition database.

5. The method of image processing according to claim 4, wherein

each element distribution p(x;

θ

_j)is a probability distribution having a single peak, and

the likelihood is calculated except for a case where a distance between a position of the peak and the positional coordinate value of the pixel exceeds a predetermined threshold value.

10. The method of image processing according to claim 4, wherein the expression (1) is mixed Gaussian distributions.

11. A method for producing a recognition database, comprising:

using a computer to perform the steps of;

(a) loading single two-dimensional-image data including a plurality of regions, each showing a prepared object,

(b) converting pixels of two-dimensional-image data into N-dimensional converted value, each N-dimensional converted value comprising a positional coordinate value of a corresponding pixel of the two-dimensional-image data and a pixel value of the corresponding pixel of the two-dimensional-image data,

(c) calculating N-dimensional estimated feature values for each prepared object which are parameters for expressing a group of N-dimensional converted values for each prepared object by the following expression,

$\begin{array}{cc}\underset{j=1}{\overset{m}{\sum <br><br>}}<br><br><br><br>{\mathrm{\xi <br><br>}}_j<br><br>p<br><br>\left(x;<br><br>{\mathrm{\theta <br><br>}}_j\right)& \left(1\right)\end{array}$

where;

x denotes the N-dimensional converted value;

ξ

_j and θ

_j are included in the N-dimensional estimated feature values;

p(x;

θ

_j) represents an element distribution;

N is an integer larger than 2; and

m is a natural number,

wherein the calculated N-dimensional estimated feature values maximize a likelihood that the group of N-dimensional converted values is expressed by the expression (1), and

(d) storing a plurality of different prepared objects in the recognition database, each prepared object including the N-dimensional estimated feature values and identifying information indicating each different prepared object in the recognition database.

12. A computer program executable by a computer program, the computer program comprising instructions that cause a computer to access a recognition database that holds a plurality of different prepared objects, each prepared object including object identifying information for identifying each prepared object and N-dimensional estimated feature values for each of the prepared object, wherein N-dimensional estimated feature values are parameters for expressing an N-dimensional converted value with plural N-dimensional regions as bases, the N-dimensional converted value is a value converted pixels of a single two-dimensional-image data of the prepared objects and the N is an integer larger than 2,

the computer program further comprises instructions for:

(a) loading a single two-dimensional-image data including an aimed object;

the single two-dimensional-image data being obtained by photo capturing an aimed object,

(b) converting pixels of the single two-dimensional-image data being obtained by photo capturing an object into N-dimensional converted values, each N-dimensional converted value comprising a positional coordinate value of a corresponding pixel of the two-dimensional-image data and a pixel value of the corresponding pixel of the two-dimensional-image data;

(c) calculating N-dimensional estimated feature values which are parameters for expressing a group of N-dimensional converted values with plural N-dimensional regions as bases, to form a prepared object by the following expression;

$\begin{array}{cc}\underset{j=1}{\overset{m}{\sum <br><br>}}<br><br><br><br>{\mathrm{\xi <br><br>}}_j<br><br>p<br><br>\left(x;<br><br>{\mathrm{\theta <br><br>}}_j\right)& \left(1\right)\end{array}$

where;

x denotes the N-dimensional converted value;

ξ

_j and θ

_j are included in the N-dimensional estimated feature values;

p(x;

θ

_j) represents an element distribution;

N is an integer larger than 2; and

m is a natural number,

wherein the calculated N-dimensional estimated feature values maximize a likelihood that the group of N-dimensional converted values is expressed by the expression (1);

(d) comparing the N-dimensional estimated feature values of the aimed object with the N-dimensional estimated feature values of the prepared objects stored in the recognition database to identify the aimed object; and

(e) outputting a result of the comparison.

13. A computer readable medium encoded with a computer program, the computer program comprising instructions that cause a computer to:

(a) load single two-dimensional-image data, each showing a prepared object;

(b) forming multidimensional variables from the two-dimensional image data, each multidimensional variable comprising a positional coordinate value of a corresponding pixel of the two-dimensional-image data and a pixel value of the corresponding pixel of the two-dimensional-image data, the pixels being contained in the single two-dimensional-image data;

(c) calculate multidimensional estimated feature values that are parameters for expressing a group of the multidimensional variables by the following expression;

$\begin{array}{cc}\underset{j=1}{\overset{m}{\sum <br><br>}}<br><br><br><br>{\mathrm{\xi <br><br>}}_j<br><br>p<br><br>\left(x;<br><br>{\mathrm{\theta <br><br>}}_j\right)& \left(1\right)\end{array}$

where;

x denotes the N-dimensional converted value;

ξ

_j and θ

_j are included in the N-dimensional estimated feature values;

p(x;

θ

_j) represents an element distribution;

m is a natural number,

wherein the calculated multidimensional estimated feature values maximize a likelihood that the group of multidimensional variables converted values is expressed by the expression (1), and

(d) execute a recognition process by comparing multidimensional estimated feature values for an aimed object with the multidimensional estimated feature values for the prepared objects stored in a recognition database.