Kannada handwritten and printed character separation

I want aa help to do my project for separating handwritten and printed character. can i achieve this by segmentation

Answer 1

[Image Analyst]

Yes you can. That's what segmentation is - identifying and separating separate distinct regions into a binary image that you can then label and measure, characterize, evaluate, or whatever.

See the published literature on OCR for examples: http://www.visionbib.com/bibliography/contentschar.html#OCR,%20Document%20Analysis%20and%20Character%20Recognition%20Systems

[Kishor s]

please can you provide any best code for segmentation

[Kishor s]

please send me the code of similar project ... please send to [e-mail]

[Image Analyst]

Attached is one possibility. It may or may not work for your images. It's designed to get a good threshold for handwritten historical documents with noise and non-uniform background.

[Kishor s]

Thank you..., In our project we are taking a application form have least noise and bachground is white. I needed help to segment can suggest any code for line and word segment. Matlab is new to me so can you provide some explanation about the result i can get after the word segment (means which pixel iam getting what other information)

[Image Analyst]

If your background is good, just apply a global threshold. See my Image Segmentation Tutorial: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862&sort=downloads_desc for a full demo that will walk you through the process.

[Kishor s]

I tried the code and it is working but want to what and all the information we get on labeled image and where it stored

[Kishor s]

how this is cropped. can i get each entity height width and starting pixel. where exactly it stored.

[Image Analyst]

Look where it's cropped out. That information will be there right where the cropping happens.

[Kishor s]

in that it labeles in vertically i want to do it in horizontally

[Image Analyst]

If you want the labeling done in a particular order, you're going to have to get the centroids and sort them based on the centroids.

[Kishor s]

thank you

[Kishor s]

hello i am sorted based on the bounding box but i cant crop it

[Image Analyst]

You can't sort like that because that BoundingBox is all the values of x, y, width, and height. You have to extract out each

bb = [blobMeasurements.BoundingBox];
% Get x
x = bb(1:4:end);
% Get y
y = bb(2:4:end);
% Get widths and heights
widths = bb(3:4:end);
heights = bb(4:4:end);

You might then want to pass the y into kmeans to find out which characters are in which line. Once you know that, then just march over with increasing x.

[classIndexes, clusterCenters] = kmeans(y, numberOfLines);
% Show what line the character is in
for k = 1 : length(classIndexes)
    fprintf('Blob %d is in line #%d.\n', k, classIndexes(k));
end

Answer 2

[Kishor s]

how this cropped