Image recognition system and recognition method thereof, and program

• US 7,532,745 B2
• Filed: 02/26/2003
• Issued: 05/12/2009
• Est. Priority Date: 02/27/2002
• Status: Expired due to Term

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Inoue, Akira
• Primary Examiner(s): Strege; John B
• Application Number:
  US10/373,167
• Publication Number:
  US 20030161504A1
• Time in Patent Office: 2,267 Days
• Field of Search: 382/118, 382/154, 382/284
• US Class Current: 382/118
• CPC Class Codes: G06V 10/507 Summing image-intensity values; Histogram projection analysis
• CPC Class Codes: G06V 10/7715 Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
• CPC Class Codes: G06V 40/16 Human faces, e.g. facial parts, sketches or expressions

Abstract

A task is to correctly classify an input image regardless of a fluctuation in illumination and a state of occlusion of the input image.

Input image sub-region extraction means 2 extracts a sub-region of the input image. Inter-pattern distance calculation means 3 calculates an inter-pattern distance between this sub-region and a sub-region of a registration image pre-filed in dictionary filing means 5 for each sub-region. Region distance value integration means 10 integrates the inter-pattern distances obtained for each sub-region. This is conducted for the registration image of each category. Identification means 4 finds a minimum value out of its integrated inter-pattern distances, and in the event that its minimum value is smaller than a threshold, outputs a category having its minimum distance as a recognition result.

Claims

1. An image recognition system comprising:

inter-pattern distance calculation means for calculating inter-pattern distances between an input image and registration images on each sub-region of a plurality of sub-regions;

inter-pattern distance integration means for selecting a portion of said plurality of sub-regions in an ascending order of said inter-pattern distances and integrating said inter-pattern distances of said selected sub-regions into an integrated inter-pattern distance; and

identification means for identifying said registration image with minimum integrated inter-pattern distance.

2. An image recognition system comprising:

fluctuation image generation means for converting a registration image into a plurality of registration images, each of the plurality of registration images differing in visualization due to a variation in posture and a variation in illumination;

subspace distance calculation means for calculating subspace distance values between each of a plurality of sub-regions of an input image, and a plurality of sub spaces of sub-regions of said registration images; and

identification means for identifying said input image based on said subspace distance values.

3. An image recognition system comprising:

inter-subspace distance calculation means for calculating, sub-region by sub-region, inter-subspace distance values between each of a plurality of subspaces of sub-regions of an input image that is comprised of a plurality of images, and each of a plurality of subspaces of sub-regions of registration images that are comprised of a plurality of images, the plurality of sub-regions of the registration images corresponding to said plurality of sub-regions of said input image; and

identification means for identifying said input image based on said inter-subspace distance values.

4. The image recognition system as set forth in claim 3, wherein said inter-subspace distance calculation means calculates the inter-subspace distance values between a subspace formed by K vectors that is generated from said sub-regions of said registration images, and a subspace formed by L vectors that is generated from said sub-regions of said input images.

5. An image recognition system comprising;

a first fluctuation image generation means for converting a registration image into a plurality of registration images, each differing in visualization due to a variation in posture or a variation in illumination;

a second fluctuation image generation means for converting an input image into a plurality of input images, each differing in visualization due to a variation in posture and a variation in illumination;

inter-subspace distance calculation means for calculating inter-subspace distance values between each of a plurality of sub-regions of said input images, and a plurality of sub-regions of said registration images that correspond to said sub-regions of said input images; and

identification means for identifying said input image based on said inter-subspace distance values.

6. The image recognition system as set forth in claim 5, wherein said inter-subspace distance calculation means calculates the inter-subspace distance values between a subspace formed by K vectors that is generated from said sub-regions of said registration images, and a subspace formed by L vectors that is generated from said sub-regions of said input images.

7. An image recognition method comprising using a computer to carry out the steps of:

calculating inter-pattern distance between an input image and registration images on each sub-region of a plurality of sub-regions;

integrating said inter-pattern distances for selecting a portion of said plurality of sub-regions in an ascending order of said inter-pattern distances and integrating said inter-pattern distances of said selected sub-regions into an integrated inter-pattern distance; and

identifying said registration image with minimum integrated inter-pattern distance.

8. An image recognition method comprising using a computer to carry out the steps of:

converting a registration image into a plurality of registration images, each of the plurality of registration images differing in visualization due to a variation in posture and a variation in illumination;

calculating subspace distance values between each of a plurality of sub-regions of an input image, and a plurality of subspaces of sub-regions of said registration images; and

identifying said input image based on said subspace distance values.

9. An image recognition method comprising using a computer to carry out the steps of:

calculating, sub-region by sub-region, inter-subspace distance values between each of a plurality of subspaces of sub-regions of an input image that is comprised of a plurality of images, and each of a plurality of subspaces of sub-regions of registration images that are comprised of a plurality of images, the plurality of sub-regions of the registration images corresponding to said plurality of sub-regions of said input image; and

identifying said input image based on said inter-subspace distance values.

10. The image recognition method as set forth in claim 9, wherein said calculating the inter-subspace distance values comprises calculating the inter-subspace distance values between a subspace formed by K vectors that is generated from said sub-regions of said registration images, and a subspace formed by L vectors that is generated from said sub-regions of said input images.

11. An image recognition method comprising using a computer to carry out the steps of:

converting a registration image into a plurality of registration images, each differing in visualization due to a variation in posture or a variation in illumination;

converting an input image into a plurality of input images, each differing in visualization due to a variation in posture and a variation in illumination;

calculating inter-subspace distance values between each of a plurality of sub-regions of said input images, and a plurality of sub-regions of said registration images that correspond to said sub-regions of said input images; and

identifying said input image based on said inter-subspace distance values.

12. The image recognition method as set forth in claim 11, wherein said calculating the inter-subspace distance values comprises calculating the inter-subspace distance values between a subspace formed by K vectors that is generated from said sub-regions of said registration images, and a subspace formed by L vectors that is generated from said sub-regions of said input images.

13. A computer readable storage medium storing a program for executing an image recognition method, the method comprising:

calculating inter-pattern distance between an input image and registration images on each sub-region of a plurality of sub-regions;

integrating said inter-pattern distances for selecting a portion of said plurality of sub-regions in an ascending order of said inter-pattern distances and integrating said inter-pattern distances of said selected sub-regions into an integrated inter-pattern distance; and

identifying said registration image with minimum integrated inter-pattern distance.

14. A computer readable storage medium storing a program for executing an image recognition method, the method comprising:

converting a registration image into a plurality of registration images, each of the plurality of registration images differing in visualization due to a variation in posture and a variation in illumination;

calculating subspace distance values between each of a plurality of sub-regions of an input image, and a plurality of subspaces of sub-regions of said registration images; and

identifying said input image based on said subspace distance values.

15. A computer readable storage medium storing a program for executing an image recognition method, the method comprising:

calculating, sub-region by sub-region, inter-subspace distance values between each of a plurality of subspaces of sub-regions of an input image that is comprised of a plurality of images, and each of a plurality of subspaces of sub-regions of registration images that are comprised of a plurality of images, the plurality of sub-regions of the registration images corresponding to said plurality of sub-regions of said input image; and

identifying said input image based on said inter-subspace distance values.

16. The computer readable storage medium as set forth in claim 15, wherein said calculating the inter-subspace distance values comprises calculating the inter-subspace distance values between a subspace formed by K vectors that is generated from said sub-regions of said registration images, and a subspace formed by L vectors that is generated from said sub-regions of said input images.

17. A computer readable storage medium storing a program for executing an image recognition method, the method comprising;

converting a registration image into a plurality of registration images, each differing in visualization due to a variation in posture and a variation in illumination;

converting an input image into a plurality of input images, each differing in visualization due to a variation in posture or a variation in illumination;

calculating inter-subspace distance values between each of a plurality of sub-regions of said input images, and a plurality of sub-regions of said registration images that correspond to said sub-regions of said input images; and

identifying said input image based on said inter-subspace distance values.

18. The computer readable storage medium as set forth in claim 17, wherein said calculating the inter-subspace distance values comprises calculating the inter-subspace distance values between a subspace formed by K vectors that is generated from said sub-regions of said registration images, and a subspace formed by L vectors that is generated from said sub-regions of said input images.

19. An image recognition system comprising:

fluctuation image generation means for converting a registration image into a plurality of registration images, each of the plurality of registration images differing in visualization due to a variation in posture or a variation in illumination;

subspace distance calculation means for calculating subspace distance values between each of a plurality of sub-regions of an input image, and a plurality of subspaces of sub-regions of said registration images; and

identification means for identifying said input image based on said subspace distance values.

20. An image recognition method comprising using a computer to carry out the steps of:

converting a registration image into a plurality of registration images, each of the plurality of registration images differing in visualization due to a variation in posture or a variation in illumination;

calculating subspace distance values between each of a plurality of sub-regions of an input image, and a plurality of subspaces of sub-regions of said registration images; and

identifying said input image based on said subspace distance values.

21. A computer readable storage medium storing a program for executing an image recognition method, the method comprising:

converting a registration image into a plurality of registration images, each of the plurality of registration images differing in visualization due to a variation in posture or a variation in illumination;

calculating subspace distance values between each of a plurality of sub-regions of an input image, and a plurality of subspaces of sub-regions of said registration images; and

identifying said input image based on said subspace distance values.