how to segment words in an image containing paragraph?

In the image,now line segmentation is done! How to segment words?? Pls help me with the code.. Thank you

Answer 1

[Image Analyst]

For each line, do the same process as you did for the lines. Sum the image vertically and look for gaps.

[sindhu c]

Thank you for the reply. Its your code only which i took from other link.. The code is attached below for line segmentation. What changes should be done for word segmentation?

[sindhu c]

After summing vertically, i.e verticalProfile = sum(thisLine,2); what should be done?

[Image Analyst]

No - you're supposed to get the horizontal profile, NOT the vertical profile.

horizontalProfile = mean(thisLine, 1);
% Columns in a word will have a sum that's pretty low
darkColumns = horizontalProfile < 210; % or whatever works.

Now wherever darkColumns is true will be a column that has letters in it. You can find individual stretches of dark columns with bwlabel

labeledColumns = bwlabel(darkColumns);

Not sure what you want to do after this. If you want to extract separate words, use ismember(), for example to get the third word

word3 = ismember(labeledColumns, 3); These are logical columns
% Extract only those columns from word #3 from the sentence band:
wordNumber3 = thisLine(:, word3); % wordNumber3 is a 2D image.

[sindhu c]

Thank you for the reply.. After segmenting i have to get segmented words as i got lines.. I have attached the code.Please check whether its correct or not.After following the instructions which you gave,nothing is being displayed for word segmentation

[sindhu c]

for clear understanding, i have got the segmented line from previous code.Now how to segment words from this line? Please help me with code

Answer 2

[Divya]

Hi Image Analyst,

Thank you for the code to segment lines. But I'm unable to segment with this code.

For the same image given above, the statement "darkColumns = horizontalProfile < 210; " gives zero column.

And when labeling such column doesn't give any result.

I tried changing threshold values. But I am getting stuck in labeling the columns. Can you please help me