Image processing method, apparatus and storage medium

• US 20040131235A1
• Filed: 11/20/2003
• Published: 07/08/2004
• Est. Priority Date: 12/13/2002
• Status: Active Grant

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Ishida, Yoshihiro, Ji, Xin, Wang, Libing, Chen, Xinwu
• Granted Patent: US 7,415,137 B2
• Time in Patent Office: Days
• US Class Current: 382/115
• CPC Class Codes: G06V 40/161 Detection; Localisation; Normalisation

Abstract

The present invention provides a method of processing an image, characterized by comprising steps of: identifying one candidate for human face region within said image; calculating a probability that said candidate for human face region represents a human face; and saving said probability as attached information to said image. The present invention also provides a method of processing an image, characterized by comprising steps of: identifying one candidate for human face region within said image; calculating a probability that said candidate for human face region represents a human face; judging whether or not said candidate for human face region represents a human face by comparing said probability with a threshold; and saving a result of said step of judging as attached information to said image. According to these methods, results of identifying candidates for human face regions will be saved to the image, and further processes to be conducted on the image will be facilitated.

Claims

1. A method of processing an image, characterized by comprising steps of:

identifying one candidate for human face region within said image;

calculating a probability that said candidate for human face region represents a human face; and

saving said probability as attached information to said image.

2. The method of processing an image according to claim 1, characterized by further comprising a step of saving identification information of said candidate for human face region as additional attached information to said image.

3. The method of processing an image according to claim 2, characterized in that said attached information and said additional attached information are saved in a predetermined format to the header file or the footer file of said image.

7. The method of processing an image according to any one of claim 1, characterized in that said step of calculating a probability is performed using K image processing apparatuses that have been trained with K training methods, wherein K is an integer greater than one and each of said K image processing apparatuses adopts a corresponding predetermined algorithm to generate M-dimensional vectors, and that

said step of calculating a probability comprises steps of:

obtaining K intermediate probabilities, p1, p2, . . . , pK, for said candidate for human face region by performing following steps within each of said K image processing apparatuses;

inputting data for said candidate for human face region;

generating a M-dimensional vector for said candidate for human face region by applying a corresponding predetermined algorithm to the input data for said candidate for human face region;

identifying among

$\underset{j=1}{\overset{M}{\prod <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{Kj}$

subspaces a subspace in which said M-dimensional vector for said candidate for human face region is located; and

assigning the value of the probability for the identified subspace to the probability for said candidate for human face region; and

calculating probability for said candidate for human face region with the following equation;

$p=\mathrm{\alpha <br><br>}\left(1-\underset{i=1}{\overset{K}{\prod <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>\left(1-\mathrm{pi}\right)\right)$

wherein α

is a factor that is less than but very closed to one.

8. The method of processing an image according to claim 7, characterized in that each of said K training methods comprises steps of:

inputting data for N image regions into said image processing apparatus, wherein N is an integer greater than one, and some of the N image regions represent human faces;

generating N M-dimensional vectors for said N image regions by respectively applying said corresponding predetermined algorithm to the input data for each of said N image regions, wherein some of the N M-dimensional vectors correspond to human faces, and the N M-dimensional vectors are distributed in a M-dimensional space;

dividing said M-dimensional space into

$\underset{j=1}{\overset{M}{\prod <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{Kj}$

subspaces, such that the number of the M-dimensional vectors distributed in each subspace is the same and equal to

$N/\underset{j=1}{\overset{M}{\prod <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{Kj},$

wherein K1, K2, . . . , KM are integers greater than one;

calculating probability for each subspace by dividing the total number of the M-dimensional vectors distributed in each subspace by the number of the M-dimensional vectors that are distributed in the same subspace and correspond to human faces; and

saving the positions and probabilities for said

$\underset{j=1}{\overset{M}{\prod <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{Kj}$

subspaces in said image processing apparatus.

9. The method of processing an image according to claim 8, characterized in that K equals to one and α

equals to one.

4. A method of processing an image, characterized by comprising steps of:

identifying one candidate for human face region within said image;

calculating a probability that said candidate for human face region represents a human face;

judging whether or not said candidate for human face region represents a human face by comparing said probability with a threshold; and

saving a result of said step of judging as attached information to said image.

5. The method of processing an image according to claim 4, characterized by further comprising a step of saving identification information of said candidate for human face region as additional attached information to said image.

6. The method of processing an image according to claim 5, characterized in that said attached information and said additional attached information are saved in a predetermined format to the header file or the footer file of said image.

10. An apparatus for processing an image, characterized by comprising:

a candidate selector for identifying one candidate for human face region within said image;

a probability calculator for calculating a probability that said candidate for human face region represents a human face; and

a probability writer for writing said probability as attached information to said image.

11. An apparatus for processing an image, characterized by comprising:

a candidate selector for identifying one candidate for human face region within said image;

a probability calculator for calculating a probability that said candidate for human face region represents a human face; and

a judging unit for judging whether said candidate for human face region represents a human face by comparing said probability with a threshold; and

a judging result writer for writing an output of said judging unit as attached information to said image.

12. A method of processing an image, characterized by comprising steps of:

obtaining at least one probability, each of said at least one probability representing possibility that one candidate for human face region within said image represents a human face;

calculating a portrait probability for said image based on said at least one probability; and

saving said portrait probability as attached information to said image.

13. The method of processing an image according to claim 12, characterized in that said attached information is saved in a predetermined format to the header file or the footer file of said image.

16. The method of processing an image according to any one of claim 12, characterized in that said step of calculating a portrait probability is performed in the following formula:

17. The method of processing an image according to any one of claim 12, characterized in that said step of calculating a portrait probability is performed in the following formula:

18. The method of processing an image according to any one of claim 12, characterized in that said step of calculating a portrait probability is performed in the following formula:

P=Max{p1, p2, p3, . . . pN}

wherein, pi denotes the probability for the i-th candidate for human face region;

N denotes the number of candidates for human face region within said image;

P denotes said portrait probability of said image.

14. A method of processing an image, characterized by comprising steps of:

obtaining at least one probability, each of said at least one probability representing possibility that one candidate for human face region within said image represents a human face;

calculating a portrait probability for said image based on said at least one probability; and

judging whether or not said image is a portrait by comparing said portrait probability with a threshold; and

saving a result of said step of judging as attached information to said image.

15. The method of processing an image according to claim 14, characterized in that said attached information is saved in a predetermined format to the header file or the footer file of said image.

19. An apparatus for processing an image, characterized by comprising:

a probability for candidate for human face region calculator, for obtaining at least one probability, each of said at least one probability representing possibility that one candidate for human face region within said image represents a human face;

a portrait probability calculator for calculating a portrait probability of said image that said image is portrait; and

a probability writer for writing said probability as attached information to said image.

20. An apparatus for processing an image, characterized by comprising:

a probability for candidate for human face region calculator, for obtaining at least one probability, each of said at least one probability representing possibility that one candidate for human face region within said image represents a human face;

a portrait probability calculator for calculating a probability that said candidate for human face region represents a human face; and

a judging unit for judging whether said image is portrait by comparing said probability with a threshold; and

a judging result writer for writing an output of said judging unit as attached information to said image.

21. A storage medium storing a computer program for processing an image, characterized by comprising:

code means for identifying one candidate for human face region within said image;

code means for calculating a probability that said candidate for human face region represents a human face; and

code means for saving said probability as attached information to said image.

22. A storage medium storing a computer program for processing an image, characterized by comprising:

code means for identifying one candidate for human face region within said image;

code means for calculating a probability that said candidate for human face region represents a human face;

code means for judging whether or not said candidate for human face region represents a human face by comparing said probability with a threshold; and

code means for saving a result of said step of judging as attached information to said image.

23. A storage medium storing a computer program for processing an image, characterized by comprising:

code means for obtaining at least one probability, each of said at least one probability representing possibility that one candidate for human face region within said image represents a human face;

code means for calculating a portrait probability for said image based on said at least one probability; and

code means for saving said portrait probability as attached information to said image.

24. A storage medium storing a computer program for processing an image, characterized by comprising:

code means for obtaining at least one probability, each of said at least one probability representing possibility that one candidate for human face region within said image represents a human face;

code means for calculating a portrait probability for said image based on said at least one probability; and

code means for judging whether or not said image is a portrait by comparing said portrait probability with a threshold; and

code means for saving a result of said step of judging as attached information to said image.