Apparatus, method and non-transitory computer readable medium to perform image recognizing using dictionary data

• US 8,611,645 B2
• Filed: 09/09/2011
• Issued: 12/17/2013
• Est. Priority Date: 03/25/2011
• Status: Active Grant

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Sumiyoshi, Masato, Nishiyama, Manabu, Watanabe, Tomoki
• Primary Examiner(s): Patel, Nirav G
• Application Number:
  US13/229,625
• Publication Number:
  US 20120243778A1
• Time in Patent Office: 830 Days
• Field of Search: None
• US Class Current: 382/159
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries

Abstract

An image recognizing apparatus includes a dictionary memory, a block determining module and a recognizing module. The dictionary memory stores dictionary data. The block determining module determines that a target block comprising a target pixel to be processed of a plurality of pixels in image data is a shared block to which the dictionary data is used or a mirror block to which the dictionary data to the shared block is used, based on a position of the target block. The recognizing module uses common dictionary data for the shared block and the mirror block, and recognizes a characteristic portion of the image expressed by the image data.

Claims

1. An image recognizing apparatus comprising:

a dictionary memory configured to store first dictionary data and second dictionary data different from the first dictionary data;

a block determining module configured to determine that a target block comprising a target pixel to be processed of a plurality of pixels in image data is a shared block to which the second dictionary data is used, a mirror block to which the second dictionary data is used, or an unshared block to which the first dictionary data is used, based on a position of the target block; and

a recognizing module configured to use the first dictionary data for the unshared block and the second dictionary data for the shared block and the mirror block, and recognize characteristic portion of the image expressed by the image data.

2. The apparatus of claim 1, further comprising a shift register configured to store gradient information indicating a gradient direction of each pixel, wherein

the block determining module generates a first dictionary address on the dictionary memory in which the first dictionary data is stored when the target block is the unshared block, and generates a second dictionary address on the dictionary memory in which the second dictionary data is stored when the target block is the shared block or the mirror block,

the shift register generates a first reference address on the dictionary memory in which a weight of the first dictionary data is stored by using the first dictionary address and generates a second reference address on the dictionary memory in which a weight of the second dictionary data is stored by using the second dictionary address, and

the recognizing module accumulates the weights stored in the first reference address and the second reference address to recognize the characteristic portion.

3. The apparatus of claim 2, wherein the shift register generates the first reference address in accordance with target gradient information of the target pixel, and generates the second reference address in accordance with a combination of co-occurrence gradient information of a co-occurrence pixel having a co-occurrence relationship with the target pixel and the target gradient information.

4. The apparatus of claim 3, wherein

the first reference address is a sum of the first dictionary address and the target gradient information or a sum of the second dictionary address and the target gradient information, and

the second reference address is a sum of the first dictionary address and a constant in accordance with the combination or a sum of the second dictionary address and the constant.

5. The apparatus of claim 3, wherein the shift register uses a reference table indicating a relationship of the target gradient information, the co-occurrence gradient information and the second reference address to generate the second reference address.

6. The apparatus of claim 1, further comprising:

a first characteristic value calculator configured to calculate a first characteristic value based on the pixels in the target block when the target block is the shared block;

an inverting module configured to invert coordinates of the pixels in the target block to generate an inverted block when the target block is the shared block; and

a second characteristic value calculator configured to calculate a second characteristic value based on the pixels in the inverted block, wherein

the recognizing module uses the first characteristic value and the second characteristic value to recognize the characteristic portion.

7. A method for recognizing a characteristic portion of an image with a dictionary memory storing first dictionary data and second dictionary data different from the first dictionary data, the method comprising:

determining that a target block comprising a target pixel to be processed of a plurality of pixels in image data is a shared block to which the second dictionary data is used, a mirror block to which the second dictionary data is used or a unshared block to which the first dictionary data is used, based on a position of the target block; and

recognizing the characteristic portion by using the first dictionary data for the unshared block and the second dictionary data for the shared block and the mirror block.

8. The method of claim 7, further comprising:

generating a first dictionary address on the dictionary memory in which the first dictionary data is stored when the target block is the unshared block;

generating a second dictionary address on the dictionary memory in which the second dictionary data is stored when the target block is the shared block or the mirror block;

generating a first reference address on the dictionary memory in which a weight of the first dictionary data is stored by using the first dictionary address; and

generating a second reference address on the dictionary memory in which a weight of the second dictionary data is stored by using the second dictionary address, wherein

in recognizing the characteristic portion, the weights stored in the first reference address and the second reference address are accumulated.

9. The method of claim 8, wherein

in generating the first reference address, the first reference address in accordance with target gradient information of the target pixel is generated, and

in generating the second reference address, the second reference address in accordance with a combination of co-occurrence gradient information of a co-occurrence pixel having a co-occurrence relationship with the target pixel and the target gradient information is generated.

10. The method of claim 9, wherein

the first reference address is a sum of the first dictionary address and the target gradient information or a sum of the second dictionary address and the target gradient information, and

the second reference address is a sum of the first dictionary address and a constant in accordance with the combination or a sum of the second dictionary address and the constant.

11. The method of claim 10, wherein in generating the second reference address, a reference table indicating a relationship of the target gradient information, the co-occurrence gradient information and the second reference address is used.

12. The method of claim 7, further comprising:

calculating a first characteristic value based on the pixels in the target block when the target block is the shared block;

inverting coordinates of the pixels in the target block to generate an inverted block when the target block is the shared block; and

calculating a second characteristic value based on the pixels in the inverted block, wherein

in recognizing the characteristic portion, the first characteristic value and the second characteristic value are used.

13. A non-transitory computer readable medium configured to store a program for recognizing a characteristic portion of an image with a dictionary memory storing first dictionary data and second dictionary data different from the first dictionary data, the program comprising instructions for:

determining that a target block comprising a target pixel to be processed of a plurality of pixels in image data is a shared block to which the second dictionary data is used, a mirror block to which the second dictionary data is used or a unshared block to which the first dictionary data is used, based on a position of the target block; and

recognizing the characteristic portion by using the first dictionary data for the unshared block and the second dictionary data for the shared block and the mirror block.

14. The medium of claim 13, further comprising:

generating a first dictionary address on the dictionary memory in which the first dictionary data is stored when the target block is the unshared block;

generating a second dictionary address on the dictionary memory in which the second dictionary data is stored when the target block is the shared block or the mirror block;

generating a first reference address on the dictionary memory in which a weight of the first dictionary data is stored by using the first dictionary address; and

generating a second reference address on the dictionary memory in which a weight of the second dictionary data is stored by using the second dictionary address, wherein

in recognizing the characteristic portion, the weights stored in the first reference address and the second reference address are accumulated.

15. The medium of claim 14, wherein

in generating the first reference address, the first reference address in accordance with target gradient information of the target pixel is generated, and

in generating the second reference address, the second reference address in accordance with a combination of co-occurrence gradient information of a co-occurrence pixel having a co-occurrence relationship with the target pixel and the target gradient information is generated.

16. The medium of claim 15, wherein

the first reference address is a sum of the first dictionary address and the target gradient information or a sum of the second dictionary address and the target gradient information, and

the second reference address is a sum of the first dictionary address and a constant in accordance with the combination or a sum of the second dictionary address and the constant.

17. The medium of claim 16, wherein in generating the second reference address, a reference table indicating a relationship of the target gradient information, the co-occurrence gradient information and the second reference address is used.