Method and device for generating image

Patent metadata

Field	Value
Publication number	US-10839244-B2
Application number	US-201816109683-A
Country	US
Kind code	B2
Filing date	Aug 22, 2018
Priority date	Aug 25, 2017
Publication date	Nov 17, 2020
Grant date	Nov 17, 2020

Abstract

The present disclosure provides a method and a device for generating an image. The method includes: a, obtaining a character recognition result corresponding to a first image, the character recognition result including one or more characters and a first confidence of each character; b, determining a second confidence of a character set including at least one of the one or more characters according to the first confidence of each character in the character set; c, determining a refined character set corresponding to the first image based on the second confidence; and d, performing image processing on a sub image corresponding to the refined character set in the first image, to obtain a second image, an annotation text corresponding to the second image including the refined character set.

First claim

What is claimed is:

1. A computer-implemented method for generating an image, comprising:
a, obtaining, by one or more computing devices, a character recognition result corresponding to a first image, the character recognition result comprising one or more characters and a first confidence of each character;
b, determining, by the one or more computing devices, a second confidence of a character set comprising at least one of the one or more characters according to the first confidence of each character in the character set;
c, determining, by the one or more computing devices, a refined character set corresponding to the first image based on the second confidence; and
d, performing, by the one or more computing devices, image processing on a sub image corresponding to the refined character set in the first image, to obtain a second image, and annotating the second image with an annotation text comprising the refined character set;
wherein, act c comprises:
sorting, by the one or more computing devices, a plurality of character sets in a descending order of the second confidence; and
determining, by the one or more computing devices, the first N character sets in the order as refined character sets corresponding to the first image, N being a natural number.

2. The method according to claim 1 , wherein, act d comprises:
determining, by the one or more computing devices, the sub image corresponding to the refined character set in the first image according to area position information of the one or more characters in the refined character set; and
performing, by the one or more computing devices, the image processing on the sub image, to obtain the second image, the annotation text corresponding to the second image comprising the refined character set.

3. The method according to claim 2 , wherein, the character recognition result further comprises the area position information.

4. The method according to claim 1 , wherein, act c comprises:
determining, by the one or more computing devices, the character set as the refined character set corresponding to the first image when the second confidence is equal to or greater than a set confidence threshold.

5. The method according to claim 1 , wherein, the character set comprises at least one of:
one text line of the one or more characters;
a plurality of text lines of the one or more characters;
a part of one text line of the one or more characters; and
different text lines of the one or more characters.

6. The method according to claim 1 , wherein, act d comprises:
performing, by the one or more computing devices, different image processing on the sub image corresponding to the refined character set in the first image, to obtain a plurality of second images, the annotation text corresponding to each second image comprising the refined character set.

7. The method according to claim 1 , wherein, act d comprises:
performing, by the one or more computing devices, the image processing on the sub image corresponding to at least one of a plurality of refined character sets in the first image in responding to existing the plurality of refined character sets, to obtain the second image, the annotation text corresponding to the second image comprising the refined character set.

8. The method according to claim 1 , wherein, the image processing comprises at least one of:
an image angle rotation processing;
an image blur processing;
an image color reversing processing;
an image zooming processing; and
superposing check noise in the image.

9. The method according to claim 1 , wherein, act c comprises:
perform, by the one or more computing devices, character recognition on the first image, to obtain a first character recognition result of the first image, the first character recognition result comprising the one or more characters, a character position of each character and the first confidence of each character; and
performing, by the one or more computing devices, combining recognition on two or more neighboring characters in the first image to obtain a second character recognition result of the first image, the second character recognition result comprising the two or more neighboring characters, a character position of the two or more neighboring characters and a confidence corresponding to the two or more neighboring characters;
obtaining, by the one or more computing devices, the character recognition result corresponding to the first image, the character recognition result comprising the first character recognition result and the second character recognition result.

10. The method according to claim 9 , wherein, act b comprises:
determining, by the one or more computing devices, whether characters corresponding to the same character position are consistent in the first character recognition result and the second character recognition result;
b1, determining, by the one or more computing devices, the second confidence of the character set according to whether the characters corresponding to the same character position are consistent and by combining the confidences of each character in the character set in the first character recognition result and the second character recognition result, the character set comprising at least one of the one or more characters.

11. The method according to claim 10 , wherein, act b1 comprises:
determining, by the one or more computing devices, the second confidence of the character set by combining the confidences of each character in the character set in the first character recognition result and the second character recognition result, when the characters corresponding to the same character position are consistent, the character set comprising at least one of the one or more characters.

12. The method according to claim 10 , wherein, act b1 comprises:
setting, by the one or more computing devices, the second confidence of the character set to zero when the characters corresponding to the same character position are not consistent, the character set comprising at least one of the one or more characters.

13. A non-transitory computer-readable storage medium having stored computer codes, wherein when the computer codes are executed, a method is executed, the method comprising:
a, obtaining a character recognition result corresponding to a first image, the character recognition result comprising one or more characters and a first confidence of each character;
b, determining a second confidence of a character set comprising at least one of the one or more characters according to the first confidence of each character in the character set;
c, determining a refined character set corresponding to the first image based on the second confidence; and
d, performing image processing on a sub image corresponding to the refined character set in the first image, to obtain a second image, and annotating the second image with an annotation text comprising the refined character set;
wherein, act c comprises:
sorting a plurality of character sets in a descending order of the second confidence; and
determining the first N character sets in the order as refined character sets corresponding to the first image, N being a natural number.

14. The computer-readable storage medium according to claim 13 , wherein, act d comprises:
determining the sub image corresponding to the refined character set in the first image according to area position information of the one or more characters in the refined character set; and
performing the image processing on the sub image, to obtain the second image, the annotation text corresponding to the second ima

Assignees

• Baidu online network technology beijing co ltd

Inventors

• Xie Shufu
• Liu Shanshan

Classifications

• G06F16/951 Primary

  Indexing; Web crawling techniques · CPC title

• G06V10/20

  Image preprocessing · CPC title

• G06V30/158

  using character size, text spacings or pitch estimation · CPC title

• G06V30/10

  Character recognition · CPC title

• G06V30/287

  of Kanji, Hiragana or Katakana characters · CPC title