Detecting text using stroke width based text detection

• US 9,235,759 B2
• Filed: 11/26/2014
• Issued: 01/12/2016
• Est. Priority Date: 05/19/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Epshtein, Boris, Ofek, Eyal, Wexler, Yonatan
• Primary Examiner(s): Park, Chan
• Assistant Examiner(s): YEH, EUENG NAN
• Application Number:
  US14/554,594
• Publication Number:
  US 20150078664A1
• Time in Patent Office: 412 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06V 20/63 Scene text, e.g. street names
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/414 Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text

Abstract

Detecting text using stroke width based text detection. As a part of the text detection, a representation of an image is generated that includes pixels that are associated with the stroke widths of components of the image. Connected components of the image are identified by filtering out portions of the pixels using metrics related to stroke width. Text is detected in the image based on the identified connected components.

Claims

1. A method of detecting text using stroke width based text detection, comprising:

generating a representation of an image, the representation including pixels that are associated with stroke widths of components of the image;

identifying connected components of the image by filtering out portions of the pixels, by applying metrics related to stroke width, so as to eliminate portions that are at least unlikely be associated with text; and

detecting text of the image, based on results of the identifying, by building a multi-nodal graph that is a string of nodes each of which is connected to all other nodes and severing connections between nodes that do not belong to a same line of text.

2. The method of claim 1, wherein the severing connections results in a number of connected sub-graphs each of which represents a plausible line of text.

3. The method of claim 1, wherein said image is selected from the group that includes natural image, infra red image, CT scan and radar.

4. The method of claim 3, wherein said image is based on vector data.

5. The method of claim 3, wherein said image is presented on a non-regular grid.

6. The method of claim 1, wherein said filtering comprises eliminating components that cannot be recognized or are not considered associated with valid text.

7. The method of claim 1, wherein said filtering comprises determining whether adjacent pixels are associated with the same stroke width.

8. The method of claim 1, wherein said filtering comprises eliminating components with an associated stroke width whose standard deviation exceeds a predetermined threshold.

9. A stroke width based text detection system, comprising

a component that directs stroke width based text detection of text in a received image, the component including:

a stroke width transform generator that generates a representation of the received image, the representation including pixels that are associated with stroke widths of components of the received image;

a connected component identifier that identifies connected components of the received image by filtering out portions of the pixels, by applying metrics related to stroke width, so as to eliminate portions that are at least unlikely be associated with text; and

a text identifier that detects text of the received image, based on results of the identifying, by building a multi-nodal graph that is a string of nodes each of which is connected to all other nodes and severing connections between nodes that do not belong to a same line of text.

10. The system of claim 9, wherein the severing connections by the text identifier results in a number of connected sub-graphs each of which represents a line of text candidate.

11. The system of claim 9, wherein said image is selected from the group that includes natural image, infra red image, CT scan and radar.

12. The system of claim 11, wherein said image is based on vector data.

13. The system of claim 11, wherein said image is presented on a non-regular grid.

14. The system of claim 9, wherein the connected component identifier filters by eliminating components that cannot be recognized or are not considered associated with valid text.

15. The system of claim 9, wherein the connected component identifier filters by determining whether adjacent pixels are associated with the same stroke width.

16. The system of claim 9, wherein the connected component identifier filters by eliminating components with an associated stroke width whose standard deviation exceeds a predetermined threshold.

17. A computer-usable storage device storing processing instructions that, when executed by one or more processors, cause the performance of a method, the method comprising:

generating a representation of an image, the representation including pixels that are associated with stroke widths of components of the image;

a connected component identifier that identifies connected components of the image by filtering out portions of the pixels, by applying metrics related to stroke width, so as to eliminate portions that are at least unlikely be associated with text; and

detecting text of the image, based on results of the identifying, by building a multi-nodal graph that is a string of nodes each of which is connected to all other nodes and severing connections between nodes that do not belong to a same line of text.

18. The storage device of claim 17, wherein the severing connections results in a number of connected sub-graphs each of which represents a suspected line of text.

19. A system, comprising

a component that directs stroke width based text detection of text in a captured image of a scene, the component including:

a stroke width transform generator that generates a representation of the captured image, the representation including pixels that are associated with stroke widths of components of the captured image;

identifying connected components of the captured image by filtering out portions of the pixels using metrics related to stroke width to exclude the portions of the pixels; and

a text identifier that detects detecting text in the captured image by building a graph that contains nodes that represent identified connected components,

wherein, initially, the graph is an intact string of nodes in which each node is connected to all other nodes, and

wherein connections between nodes that are identified as not belonging to a same text line are severed, based on;

distance in a vertical direction;

an amount of intersection in a horizontal direction; and

a difference in average stroke width between connected components.

20. The system of claim 19, wherein every text line is presumed to be horizontal, and wherein connected components in a same text line have similar average stroke width values.