I suffered from similar issues and fixed the problem by adding a line to 
textord/colfind.cpp:

Between

#endif // GRAPHICS_DISABLED

and

SetBlockRuleEdges(input_block);

I added:

input_block->noise_blobs.clear(); // remove noise blobs

This will remove noise blobs during the segmentation of blocks and prevent 
noise blobs from being added to the text block around them. I think it is a 
dirty hack, but it will probably give you better results. Maybe we have to 
tackle this problem in a more in-depth solution in the future.

Changing the constant

const double kMinMediumSizeRatio = 0.25;

to

const double kMinMediumSizeRatio = 0.15;

in blobbox.cpp also helped to improve the results. You can try to adjust 
that constant to your needs.

Paul


Am Donnerstag, 28. August 2014 10:04:33 UTC+2 schrieb 葉家忠:
>
> I use Tesseract to recognize the simplified chinese character
>
> Since some noise of the source image  can't be removed, so I decide to fix 
> the source code to remove the incorrect result.
>
> Since the each of the chinese charactor size is fix-sized, so the nose can 
> be found easily because its size will be much smaller than a normal 
> character. 
>
> I've tried to set the parameter "textord_heavy_nr" to  true to remove the 
> noise, but it won't work  because in some case it will remove some 
> importart parts of a chinese character which is quite necessary to form a 
> complete chinese character
>
> Can any one tell me how to fix the code that remove the result lastly 
> decided by Tesseract which size is smaller than specific blob size?
>
> I really thank you for helping~
>
>
>
> ps: the attached file show 3 characters but it will be recognized as 4 
> characters because of the noise. 
>

-- 
You received this message because you are subscribed to the Google Groups 
"tesseract-ocr" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/tesseract-ocr.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/tesseract-ocr/adec96da-f019-44e0-9e04-d81f9de1aa50%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to