segmentation character with character that aren't connected

hello..:), i have problem to segmentation character with character that aren’t connected like arabic character and the letter "i". i used bwlabel and regionprops function but the result is wrong. please help me

[Andreas Goser]

Please help us to help you by sending code, data, information about release, OS and a description of "wrong".

[naimah lubis]

i used matlab R2008b, this is my code:

a=imread('IMG_0002.tif');

b= bwmorph(~a,'skel',inf);

BW=bwlabeln(~a,1);

stats = regionprops(BW);

for index=1:length(stats)

if stats(index).Area > 1500 && stats(index).BoundingBox(3)*stats(index).BoundingBox(4) < 80000

x = ceil(stats(index).BoundingBox(1))

y = ceil(stats(index).BoundingBox(2))

widthX = floor(stats(index).BoundingBox(3)-1)

widthY = floor(stats(index).BoundingBox(4)-1)

subimage(index) = {BW(y:y+widthY,x:x+widthX,:)};

figure, imshow(subimage{index})

end

end

this image is ancient document with Ka-Ga-Nga character. ka-Ga-Nga is the ancient script in bengkulu(indonesia).

i will recognize the character with Learning Vector Quantization.

the result of segmentation step is not valid, many character has been cropping, should one character is divided into two or three characters. you have solution? thanks:)

[Image Analyst]

Where did you upload your picture?

http://www.mathworks.com/matlabcentral/answers/7924-where-can-i-upload-images-and-files-for-use-on-matlab-answers

[naimah lubis]

thanks for the link sir..,

this is my picture:

http://www.sendspace.com/file/jao2ve

i need your help sir..,

Answer 1

[Image Analyst]

OCR is not an easy task. You're better off buying something from companies that have worked on it for years. If they can spend decades working on it, you know it can't be easy. If you have very simple cases and are looking for a not-very-robust solution, you can try to find centroids and see how close they are to each other.

[naimah lubis]

thanks for the answer but i don't understand "try to find centroids and see how close they are to each other", how i can do that??

[Image Analyst]

See my BlobsDemo tutorial on that: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862