Separation of Overlapping and Touching Lines

Hi..

Can anyone suggest an approach or method for the separation of overlapping and touching lines within handwritten documents

I can segment text lines. Now i want to separate the connected lines Here is my image link

http://oi41.tinypic.com/24m9do5.jpg

Thanks

[Anand]

Did you use the activecontour function for this segmentation? If so, what method did you use 'Chan-Vese' or 'edge'? Also, did you try fiddling with the smoothness parameter? Try reducing the smoothness parameter. That might give you a closer segmentation.

[consortium]

The implementation of active contours is done using level sets. reducing the smoothness parameter do not give good results on different types of documents. I need post-processing to separate overlapping and touching lines

Answer 1

[Image Analyst]

Try a morphological closing with imclose().

[consortium]

I tried morphological operators. I not only separates the overlapping lines but also separate lines not overlap and here is the result with the closure operator http://oi44.tinypic.com/10h0tja.jpg

[Image Analyst]

Well just use imline() to manually break apart the connections. Here's your demo:

fontSize = 20;
%----- Burn white line into image -----
burnedImage = imread('pout.tif');
% Create line mask, h, as an ROI object over the second image in the bottom row.
subplot(2, 2, 1);
imshow(burnedImage);
title('Original Image', 'FontSize', fontSize);
axis on;
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0 0 1 1]);
message = sprintf('Click and drag out a line between the end points.');
uiwait(msgbox(message));
hLine = imline(); % Arguments 2 and 3 define line endpoints.
% Create a binary image ("mask") from the ROI object.
binaryImage2 = hLine.createMask();
% Display the line mask.
subplot(2, 2, 2);
imshow(binaryImage2);
title('Binary mask of the line', 'FontSize', fontSize);
% Burn line into image by setting it to 255 wherever the mask is true.
burnedImage(binaryImage2) = 255;
% Display the image with the "burned in" line.
subplot(2, 2, 3);
imshow(burnedImage);
title('New image with line burned into image', 'FontSize', fontSize);

[consortium]

Thanks for all but surely there is some other solution (automatic solution),

[Image Analyst]

Why? Why does it have to be automatic, rather than quick and effective? Do you want to get the job done , or do you want to spend weeks researching it (more time than you'd save with an automatic solution)? Unless you have thousands of images (which I doubt), I'd just get it done.

[consortium]

Thank you for all. I tried this method using the watershed over the distance function of the connected component set to separate them like here and here here some results obtained http://oi39.tinypic.com/adkprk.jpg and http://oi39.tinypic.com/29e6p86.jpg ... this method is inappropriate for the shapes textelines but I think I found an automated solution using histogram projection profile but isn't appropriate for all types of documents http://oi41.tinypic.com/69q99t.jpg ... I'm waiting for any sugestion