Method and apparatus for detecting texts included in a specific image

Patent metadata

Field	Value
Publication number	US-9710703-B1
Application number	US-201615211127-A
Country	US
Kind code	B1
Filing date	Jul 15, 2016
Priority date	Jul 15, 2016
Publication date	Jul 18, 2017
Grant date	Jul 18, 2017

Abstract

A method for detecting texts included in a specific image is disclosed. The method includes steps of: (a) an apparatus detecting or allowing another device to detect one or more text candidates in the specific image by referring to feature values of pixels in the specific image; (b) the apparatus classifying or allowing another device to classify one or more weak texts in the specific image as strong texts by referring to information on at least one text classified as the strong text in another image related to the specific image if more than a certain percentage of the detected text candidates are classified as the weak texts as a result of comparison between at least one threshold value and at least one feature value of at least one pixel selected in a region where the detected text candidates are included or a value converted from the feature value.

First claim

What is claimed is:

1. A method for detecting texts included in a specific image, comprising steps of:
(a) an apparatus detecting or allowing another device to detect one or more text candidates in the specific image by referring to feature values of pixels in the specific image;
(b) the apparatus classifying or allowing another device to classify one or more weak texts in the specific image as strong texts by referring to information on at least one text classified as a strong text in another image related to the specific image if more than a certain percentage of the detected text candidates are classified as the weak texts as a result of comparison between at least one threshold value and at least one feature value of at least one pixel selected in a region where the detected text candidates are included or a value converted from the feature value.

2. The method of claim 1 , wherein, at the step of (b), the apparatus determines or allows another device to determine whether the weak texts are converted into the strong texts or not by referring to the information on the text classified as the strong text in the another image related to the specific image if all the detected text candidates are classified as the weak texts.

3. The method of claim 1 , wherein, at the step of (b), (i) if the multiple text candidates detected in the specific image are adjacent to each other, and (ii) if more than the certain percentage of the multiple text candidates adjacent to each other are classified as the weak texts as the result of comparison between the at least one threshold value and the at least one feature value of the at least one pixel selected in the region where the detected multiple text candidates are included or the value converted from the feature value, the apparatus classifies or allows another device to classify the weak texts among the multiple text candidates as the strong texts by referring to the information on the text classified as the strong text in the another image related to the specific image.

4. The method of claim 1 , wherein the step of (b) includes steps of:
(b1) the apparatus (i) classifying or allowing another device to classify each of the detected text candidates as a strong text or a non-strong text by comparing the at least one feature value or the value converted from the feature value with at least a first threshold value and (ii) classifying or allowing another device to classify each of the text candidates classified as the non-strong text as a weak text or a non-text by comparing the at least one feature value or the value converted from the feature value with at least a second threshold value, and
(b2) the apparatus classifying or allowing another device to classify the weak texts in the specific image as the strong texts by referring to the information on the text classified as the strong text in the another image related to the specific image if more than the certain percentage of the detected text candidates are classified as the weak texts.

5. The method of claim 4 , wherein, at the step of (b1), if there is at least one middle threshold value between the first and the second threshold values, the apparatus classifies or allows another device to classify (i) the text candidate classified as the non-strong text as the strong text or a subordinated weak text by comparing the at least one feature value or the value converted from the feature value with the middle threshold, and (ii) the subordinated weak text as the weak text or the non-text by comparing the at least one feature value or the value converted from the feature value with the second threshold value.

6. The method of claim 1 , wherein, if the specific image is an (i)th image frame of a video, the another image related to the specific image is an (i−k)th image frame, provided that k is an integer equal to or greater than 1, wherein “i” is an integer index specifying the specific image among image frames of the video.

7. The method of claim 1 , wherein, if the specific image is an (i)th image frame of a video, the another image related to the specific image is an (i+m)th image frame, provided that m is an integer equal to or greater than 1, wherein “i” is an integer index specifying the specific image among image frames of the video.

8. The method of claim 1 , wherein, if the specific image is an (i)th image frame of a video, the another image related to the specific image includes multiple image frames located before, after, or before and after the (i)th image frame, wherein “i” is an integer index specifying the specific image among image frames of the video.

9. The method of claim 1 , after the step of (a), further comprising a step of: (a1) the apparatus identifying or allowing another device to identify at least one text candidate set including text candidates whose corresponding regions overlap by at least a specific percentage, and selecting or allowing another device to select the text candidate satisfying a predetermined condition among the identified text candidate set, and
wherein, at the step of (b), if the text candidate satisfying the predetermined condition is classified as the weak text with more than the certain percentage, the apparatus classifies or allows another device to classify the weak text in the specific image as the strong text by referring to the information on the text classified as the strong text in the another image related to the specific image.

10. The method of claim 1 , wherein, at the step of (a), the apparatus detects or allows another device to detect one or more text candidates while changing intensity levels of the pixels of a specific region in the specific image, and builds or allows another device to build a tree by allocating (i) a low-level text candidate whose intensity level in the specific region is relatively high and (ii) a text candidate whose intensity level is the highest among high-level text candidates whose intensity level in the specific region is relatively low, to a child node and a parent node, respectively, and wherein respective regions of the high-level text candidates include a region of the low-level text candidate.

11. The method of claim 10 , wherein, if there occurs no less than a specific percentage of an overlap between at least one first region where at least one text candidate corresponding to at least one child node in the tree is included and at least one second region where at least one text candidate, with a relatively low intensity level of the pixel, corresponding to at least one parent node of the child node is included for no less than a specified number of times, provided that the second region is larger than the first region and includes the first region, the apparatus selects or allows another device to select at least one text candidate corresponding to the child node if difference between an area of a region of the text candidate of the child node and that of its parent node are relatively small; and
wherein, at the step of (b), the apparatus classifies or allows another device to classify the weak text of the specific image as the strong text by referring to the information on the text classified as the strong text in the another image related to the specific image if more than the certain percentage of the detected text candidates are classified as the weak texts.

12. The method of claim 1 , further comprising a step of: (c) the apparatus grouping or allowing another device to group the strong texts if the number of the strong texts converted from the weak texts are more than two.

13. The method of claim 12 , wherein, if a first strong text and a second strong text converted from a first weak t

Assignees

• Stradvision Korea Inc
• Stradvision Inc

Inventors

• Cho Hojin

Classifications

• G06K9/4671

  Physics · mapped topic

• G06F17/21

  Physics · mapped topic

• G06K9/00442 Primary

  Physics · mapped topic

• G06V20/63 Primary

  Scene text, e.g. street names · CPC title