Text Detection Using Multi-Layer Connected Components With Histograms

• US 20130129216A1
• Filed: 11/21/2011
• Published: 05/23/2013
• Est. Priority Date: 11/21/2011
• Status: Active Grant

• Current Assignee: Nokia Technologies Oy (Nokia Corporation)
• Original Assignee: Nokia Corporation
• Inventors: Grzeszczuk, Radek, Tsai, Shang-hsuan, Parameswaran, Vasudev
• Granted Patent: US 8,611,662 B2
• Time in Patent Office: Days
• US Class Current: 382/170
• CPC Class Codes: G06V 20/62 Text, e.g. of license plates, overlay texts or captions on TV images
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/18076 by analysing connectivity, e.g. edge linking, connected component analysis or slices

Abstract

A digital image is converted to a multiple level image, and multiple scale sets are formed from connected components of the multiple level image such that different ones of the scale sets define different size spatial bins. For each of the multiple scale sets there is generated a count of connected components extracted from the respective scale set for each spatial bin; and adjacent spatial bins which represent connected components are linked. Then the connected components from the different scale sets are merged and text line detection is performed on the merged connected components. In one embodiment each of the scale sets is a histogram, and prior to linking all bins with less than a predetermined count are filtered out; and each histogram is extended such that counts of adjacent horizontal and vertical bins are added (single region bins are filtered out) and the linking is on the extended histograms.

Claims

1. An apparatus comprising:

at least one processor; and

at least one memory including computer program code,

in which the at least one memory and the computer program code are configured, with the at least one processor, to cause the apparatus at least to;

convert a digital image to a multiple level image;

form multiple scale sets from connected components of the multiple level image, in which different ones of the scale sets define different size spatial bins;

for each of the multiple scale sets;

generate a count of connected components extracted from the respective scale set for each spatial bin; and

link adjacent spatial bins which represent connected components;

merge the connected components from the different scale sets; and

perform text line detection on the merged connected components.

2. The apparatus according to claim 1, in which for each of the multiple scale sets the count of components for each spatial bin comprises a histogram, and

the at least one memory and the computer program code are configured with the at least one processor to cause the apparatus to at least further:

for each of the multiple scale sets, after generating the histogram and prior to linking the adjacent spatial bins, filter out all spatial bins in which the count is less than a predetermined value.

3. The apparatus according to claim 2, in which each histogram defines a different size bin than any other histogram.

4. The apparatus according to claim 3, in which each histogram comprises at least two overlapping grids defining same size bins.

5. The apparatus according to claim 2, in which the at least one memory and the computer program code are configured with the at least one processor to cause the apparatus to at least further:

for each of the multiple scale sets, after generating the histogram, extend the histogram by adding to each spatial bin the count from vertically and horizontally adjacent spatial bins; and

thereafter

filter out spatial bins comprising single regions;

in which the linking of the adjacent spatial bins is performed on the extended histogram after the filtering.

6. The apparatus according to claim 5, in which all the spatial bins having a count less than the predetermined threshold are filtered out from the extended histogram.

7. The apparatus according to claim 6, in which the at least one memory and the computer program code are configured with the at least one processor to cause the apparatus to further perform at least one of the following:

test text stroke width when linking adjacent bins or when performing text line detection to filter out false text character candidates; and

compare size of characters and only link adjacent spatial bins if the underlying characters are of similar size.

8. The apparatus according to claim 1, in which the components are extracted from each multiple level using maximally stable extremal regions or binarization.

9. A method comprising:

converting a digital image to a multiple level image;

form multiple scale sets from connected components of the multiple level image by at least one processor, in which different ones of the scale sets define different size spatial bins;

for each of the multiple scale sets;

generating a count of connected components extracted from the respective scale set for each spatial bin; and

linking adjacent spatial bins which represent connected components;

merging the connected components from the different scale sets; and

performing text line detection on the merged connected components.

10. The method according to claim 9, in which for each of the multiple scale sets the count of connected components for each spatial bin comprises a histogram, the method further comprising:

for each of the multiple scale sets, after generating the histogram and prior to linking the adjacent spatial bins, filtering out all spatial bins in which the count is less than a predetermined value.

11. The method according to claim 10, in which each histogram defines a different size bin than any other histogram.

12. The method according to claim 11, in which each histogram comprises at least two overlapping grids defining same size bins.

13. The method according to claim 10, the method further comprising:

for each of the multiple scale sets, after generating the histogram, extending the histogram by adding to each spatial bin the count from vertically and horizontally adjacent spatial bins; and

thereafter

filtering out spatial bins comprising single regions;

in which the linking of the adjacent spatial bins is performed on the extended histogram after the filtering.

14. The method according to claim 13, in which all the spatial bins having a count less than the predetermined threshold are filtered out from the extended histogram.

15. The method according to claim 14, the method further comprising at least one of the following:

testing text stroke width when linking adjacent bins or when performing text line detection to filter out false text character candidates; and

comparing size of characters and only linking adjacent spatial bins if the underlying characters are of similar size.

16. The method according to claim 9, in which the components are extracted from each multiple level using maximally stable extremal regions or binarization.

17. A non-transitory computer readable memory storing a program of instructions comprising:

code for converting a digital image to a multiple level image;

code for forming multiple scale sets from connected components of the multiple level image, in which different ones of the scale sets define different size spatial bins;

for each of the multiple scale sets;

code for generating a count of connected components extracted from the respective scale set for each spatial bin; and

code for linking adjacent spatial bins which represent connected components;

code for merging the connected components from the different scale sets; and

code for performing text line detection on the merged connected components.

18. The non-transitory computer readable memory according to claim 17, further comprising:

for each of the multiple scale sets, after the histogram is generated and prior to linking the adjacent spatial bins, code for filtering out all spatial bins in which the count is less than a predetermined value.

19. The non-transitory computer readable memory according to claim 18, further comprising:

for each of the multiple scale sets, after the histogram is generated, code for extending the histogram by adding to each spatial bin the count from vertically and horizontally adjacent spatial bins; and

thereafter

code for filtering out spatial bins comprising single regions;

in which the adjacent spatial bins which are linked are from the extended histogram after the filtering.

20. The non-transitory computer readable memory according to claim 19, in which all the spatial bins having a count less than the predetermined threshold are filtered out from the extended histogram.