hi ,
I have developed the line remove algorithm in the scanned document images using morphological operation but this erodes the characters in the document images. Please suggest me any paper or method for removing the lines. Below i have added the sample file and i have used image closing function for removing the lines
This is a hard question to answer unless more information is provided. It sounds like you might be using an erode algorithm to remove the lines? If so, other objects in the image (like text) will get eroded as well. If the lines are fairly straight, you might want to try using a Hough or Radon transform to locate the lines in the image. Once you have located them, you can perform operations along the lines you have identified to remove them.
If you want to stick to what you have already, you might just try using a dilation algorithm to restore what was eroded. If the erosion operation removes the lines completely, then a dilation operation should help restore some of the text as long as the text hasn't been completely removed.
Use color spaces. text is blue, lines are black
As I can see your task is to elliminate lines, and pattern of this lines is predefined. So you can find all lines, remove them, and all the rest is text, more ower you can define what text is writen in what cell.
Also you can try convlution, that elliminates vertical and horizontal lines.
Instead of using the Hough transform, you could also iteratively rotate the image and project the result on the horizontal and vertical axis. (In the example given, you would need to treat the scan as two halves...) The rotation angle leading to the sharpest peaks in the projections gets your scan straight. The location of the peaks tells you the position of the lines, and removing them is straightforward.
If the handwriting crosses the lines, hopefully there is a pixelvalue difference with parts of the lines that are not crossed. If not, you need to remember the connectivity before line removal.
Erosion will not work: the handwriting is about as thin as the lines.
In order to remove the lines from a document, and obtain a new document that appears as not having any lines (so not just replacing the lines with another colour) there are two steps:
1) detecting the lines
2) inpainting ( colouring the the pixels corresponding to the lines)
For the first task, I do agree with Nicholas Schaub and Lambert Zijp.
As for the second, you can probably take a look at these general introductions:
http://www.math.ucla.edu/~imagers/htmls/inp.html
http://math.univ-lyon1.fr/~masnou/fichiers/publications/survey.pdf
Matlab code for the most widespread method is available.
- Badges
- Science topic