CHARACTER RECOGNITION METHOD AND DEVICE

• US 20150131908A1
• Filed: 01/23/2015
• Published: 05/14/2015
• Est. Priority Date: 07/16/2013
• Status: Active Grant

• Current Assignee: Tencent Technology Shenzhen Company Limited (Tencent Holdings Limited)
• Original Assignee: Tencent Technology Shenzhen Company Limited (Tencent Holdings Limited)
• Inventors: Mei, Shuqi, Chen, Yongjun, Liu, Boxing
• Granted Patent: US 9,349,062 B2
• Time in Patent Office: Days
• US Class Current: 382/182
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• CPC Class Codes: G06V 20/62 Text, e.g. of license plates, overlay texts or captions on TV images
• CPC Class Codes: G06V 20/635 Overlay text, e.g. embedded captions in a TV program
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/162 Quantising the image signal
• CPC Class Codes: G06V 30/224 of printed characters having additional code marks or containing code marks

Abstract

A character recognition method may include at least the following steps. A location step may include acquiring an image and locating a character region of the image. The character region may include a character and a local background. The method may further include a background judgment step for determining whether the local background is a complex background; a determination step for determining a color of the character if the local background is a complex background; a construction step for constructing a mask for the character by combining the color of the character and a character region; and a first recognition step for extracting the character from the character region by using the mask, recognizing the character, and outputting the recognition result. A character recognition device is further provided.

Claims

1. A character recognition method implemented on a processor and a memory, the memory including instructions to configure the processor to perform the method comprising:

acquiring an image and locating a character region in the image by the processor, wherein the character region comprises a character and a local background;

determining, by the processor, whether the local background is a complex background;

determining, by the processor, a color of the character if the local background is a complex background;

constructing, by the processor, a mask for the character by combining the color of the character and a color of the character region;

extracting, by the processor, the character from the character region by using the mask; and

recognizing the character and outputting the recognition result by the processor.

2. The character recognition method according to claim 1, further comprising:

acquiring a position coordinate of the character region and stroke information of the character by the processor.

3. The character recognition method according to claim 2, further comprising:

generating, by the processor, a binary graph according to the position coordinate of the character region and the stroke information of the character, wherein the binary graph comprises a first pixel group comprising pixels set at a first preset gray level, and a second pixel group comprising pixels set at a second preset gray level, wherein the first pixel group represents the character in the image and the second pixel group represents a background of the image.

4. The character recognition method according to claim 3, further comprising:

performing, by the processor, an expansion process on the generated binary graph.

5. The character recognition method according to claim 4, further comprising:

determining, by the processor, pixels corresponding to the local background in the character region based on the binary graph;

determining, by the processor, statistics for each color channel of the pixels corresponding to the local background, and determining a main color of the local background;

counting, by the processor, number of the pixels corresponding to the main color of the local background;

calculating, by the processor, a first ratio comprising a ratio of the number of the pixels corresponding to the main color of the local background and a total number of the pixels of the local background; and

determining the local background is a complex background in response to the first ratio being less than a first threshold.

6. The character recognition method according to claim 5, further comprising:

comparing, by the processor, the first ratio with a second threshold in response to the first ratio not being less than the first threshold;

determining, by the processor, pixels corresponding to background of the image based on the binary graph in response to the first ratio not being larger than the second threshold;

determining, by the processor, statistics for each color channel of the pixels corresponding to the whole background, and determining the main color of the whole background;

counting, by the processor, total number of the pixels corresponding to the main color of the whole background;

calculating, by the processor, a second ratio comprising a ratio of the number of the pixels corresponding to the main color of the whole background and a total number of the pixels of the whole background; and

determining, by the processor, that the local background is a complex background in response to the second ratio being less than a third threshold.

7. The character recognition method according to claim 3, wherein further comprising:

determining, by the processor, the pixels corresponding to the character based on the binary graph;

determining, by the processor, statistics for the colors of the pixels corresponding to the character; and

determining, by the processor, a color of the character by clustering the colors of the pixels of the character.

8. The character recognition method according to claim 7, further comprising:

calculating, by the processor, a distance of the color of each pixel in the character region from the color of the character;

generating, by the processor, a color distance graph by using the distance as color value of pixels in the color distance graph; and

constructing, by the processor, the mask by binarization of the color distance graph.

9. The character recognition method according to claim 8, wherein the binarization of the color distance graph further comprises:

performing a local-adaptive binarization process on the color distance graph.

10. The character recognition method according to claim 8, wherein the constructing the mask further comprises:

removing spot noise from the mask.

11. The character recognition method according to claim 1, wherein locating the character region further comprises:

detecting an angle of inclination of the character.

12. The character recognition method according to claim 11, wherein after the construction of the mask, the method further comprises:

performing tilt correction on the character region and on the mask according to the angle of inclination.

13. The character recognition method according to claim 1, further comprising:

performing, by the processor, layout analysis and single character segmentation on the mask; and

extracting, by the processor, each character from the character region by using the mask and recognizing the each extracted character.

14. The character recognition method according to claim 1, further comprising:

performing, by the processor, an expansion process on the mask.

15. The character recognition method according to claim 1, further comprising:

normalizing the extracted character and recognizing the character by the processor.

16. The character recognition method according to claim 11, further comprising:

performing, by the processor, a binarization process on the character region in response to the local background not being a complex background; and

performing a layout analysis and single character segmentation on the binarized character region, extracting each character in the binarized character region and recognizing the each character, and outputting the recognition result by the processor.

17. The character recognition method according to claim 16, wherein after performing binarization process on the character region, the method further comprises:

performing tilt correction on the character region according to the angle of inclination.

18. A character recognition device, comprising a processor and a memory, the memory including instructions to configure the processor to comprise:

a location component configured to acquire an image, and locate a character region of the image, wherein the character region comprises a character and a local background;

a background judgment component configured to identify the local background as a complex background;

a determination component configured to determine a color of the character if the local background is identified as a complex background;

a construction component configured to construct a mask for the character by combining the color of the character and a color of the character region; and

a first recognition component configured to extract the character from the character region by using the mask, to recognize the character, and further to output the recognition result.

19. The character recognition device according to claim 18, wherein the location component is further configured to:

acquire coordinates of the character region and stroke information of the character.

20. The character recognition device according to claim 19, wherein the background judgment component is further configured to:

generate a binary graph according to coordinates of the character region and the stroke information of the character, wherein the binary graph comprises a first pixel group with pixels set at a first preset value and a second pixel group with pixels set at a second preset value, the first pixel group corresponding to the character in the image and the second pixel group corresponding to a whole background of the image.