Chinese, Japanese, or Korean language detection

1. A method for determining a text contains Chinese, Japanese, or Korean characters, the method comprising:

receiving a document image;

binarizing the document image;

searching for connected components in the binarized document image;

identifying a plurality of fragments based on the connected components;

determining a language hypothesis for each fragment of the plurality of fragments, wherein the language hypothesis has a probability rating;

selecting a subset of fragments from the plurality of fragments having highest probability ratings;

verifying, using a processor, the language hypothesis of each fragment in the subset of fragments; and

determining, using the processor, that Chinese, Japanese, or Korean (CJK) characters are present in the received document image based at least on the verification of the language hypothesis of the subset of fragments.

2. The method of claim 1 , further comprising:

analyzing features of a first subset of the plurality of fragments; and

determining a document characteristic based upon the analyzing features of the first subset of the plurality of fragments, wherein determining the language hypothesis of each fragment is based in part on the document characteristic.

3. The method of claim 2 , wherein the document characteristic is the document orientation.

4. The method of claim 2 , further comprising:

determining a second, different value of the document characteristics based upon analyzing features of a second, different subset of the plurality of fragments; and

reanalyzing the features of the first subset of the plurality of fragments using the second, different value of the document characteristics.

5. The method of claim 1 , further comprising:

recognizing the subset of fragments in each of four orientations;

calculating a recognition confidence of each of the selected subset of fragments in each of the four orientations;

determining for each of the subset of fragments a vote for an orientation based upon the calculated confidence level; and

determining an orientation of the document image based upon the votes.

6. The method of claim 1 , wherein a Bayesian network is used to determine the language hypothesis of each fragment of the plurality of fragments.

7. The method of claim 1 , wherein determining a language hypothesis of a fragment comprises identifying features of the fragment, wherein the features are based on information about raster and geometric properties of the fragment.

8. The method of claim 7 , wherein the features of the fragment comprise at least one of:

a natural logarithm of a ratio of a width and a height of the fragment, a ratio of a horizontal strokes count and the fragment height, a ratio of a vertical strokes count and the fragment width, a ratio of a maximum horizontal black stroke length and the fragment height, and a ratio of the maximum horizontal white stroke length and the fragment width.

9. The method of claim 1 , wherein a fragment comprises one of: a single character, two or more agglutinated characters, a portion of a single character, a single character and a portion of a second character.

10. The method of claim 5 , wherein each vote for a first orientation is further checked for existence of a European neighbor character in any orientation.

11. A system comprising:

one or more processors configured to:

receive a document image;

binarize the document image;

search for connected components in the binarized document image;

identify a plurality of fragments based on the connected components;

determine a language hypothesis for each fragment of the plurality of fragments, wherein the language hypothesis has a probability rating;

select a subset of fragments from the plurality of fragments having highest probability ratings;

verify the language hypothesis of each fragment in the subset of fragments; and

determine that Chinese, Japanese, or Korean (CJK) characters are present in the received document image based at least on the verification of the language hypothesis of the subset of fragments.

12. The system of claim 11 , wherein the one or more processors are further configured to:

analyze features of a first subset of the plurality of fragments; and

determine a document characteristic based upon the analyzing features of the first subset of the plurality of fragments, wherein determining the language hypothesis of each fragment is based in part on the document characteristic.

13. The system of claim 12 , wherein the document characteristic is the document orientation.

14. The system of claim 12 , wherein the one or more processors are further configured to:

determine a second, different value of the document characteristics based upon analyzing features of a second, different subset of the plurality of fragments; and

reanalyze the features of the first subset of the plurality of fragments using the second, different value of the document characteristics.

15. The system of claim 11 , wherein the one or more processors are further configured to:

recognize the subset of fragments in each of four orientations;

calculate a recognition confidence of each of the selected subset of fragments in each of the four orientations;

determine for each of the subset of fragments a vote for an orientation based upon the calculated confidence level; and

determine an orientation of the document image based upon the votes.

16. A non-transitory computer-readable medium having instructions stored thereon, the instructions comprising:

instructions to receive a document image;

instructions to binarize the document image;

instructions to search for connected components in the binarized document image;

instructions to identify a plurality of fragments based on the connected components;

instructions to determine a language hypothesis for each fragment of the plurality of fragments, wherein the language hypothesis has a probability rating;

instructions to select a subset of fragments from the plurality of fragments having highest probability ratings;

instructions to verify the language hypothesis of each fragment in the subset of fragments; and

instructions to determine that Chinese, Japanese, or Korean (CJK) characters are present in the received document image based at least on the verification of the language hypothesis of the subset of fragments.

17. The non-transitory computer readable medium of claim 16 , the instructions further comprising:

instructions to analyze features of a first subset of the plurality of fragments; and

instructions to determine a document characteristic based upon the analyzing features of the first subset of the plurality of fragments, wherein determining the language hypothesis of each fragment is based in part on the document characteristic.

18. The non-transitory computer readable medium of claim 17 , wherein the document characteristic is the document orientation.

19. The non-transitory computer readable medium of claim 17 , the instructions further comprising:

instructions to determine a second, different value of the document characteristics based upon analyzing features of a second, different subset of the plurality of fragments; and

instructions to reanalyze the features of the first subset of the plurality of fragments using the second, different value of the document characteristics.

20. The non-transitory computer-readable medium of claim 16 , the instructions further comprising:

instructions to recognize the subset of fragments in each of four orientations;

instructions to calculate a recognition confidence of each of the selected subset of fragments in each of the four orientations;

instructions to determine for each of the subset of fragments a vote for an orientation based upon the calculated confidence level; and

instructions to determine an orientation of the document image based upon the votes.