Legacy ImageMagick Discussions Archive

I'm testing ImageMagick for image preprocessing before pass them to an OCR software.
At the end of the cicle, the software I'm developing will be compiled cross platform (win/lin/mac); my software will launch ImageMagick on different platform through a shell command.
The images that ocr have to recognize contain a grid, therefore the OCR doesn't capture the text inside the grid.

After I've googled for searching a solution, I've find some syntax and now I'm trying to set them for my scope: lines grid removing.

This is the syntax on windows; it works correctly and remove vertical and horizontal lines with a lenght>100 pixel. The same syntax runned on Osx, returns an error like this:
What's wrong? How can I use this syntax on Osx?

Thanks in advance

Sincerely
Pietro

try double quotes around the whole rectangle: expression or just use escapes as \^\<

Rectangle:1x100+0+0^<

In Windows, "^" is the escape character. The Windows command interpreter sees this, and passes the next character "<" as a literal to IM, instead of interpreting it as file redirection.

You don't want "^" on non-Windows platforms. You might substitute the appropriate escape character, thus:

Rectangle:1x100+0+0\<

Sorry I missed the fact that he was starting from Windows. IM uses ^ for rescaling and in windows would need to be ^^, where appropriate. I mistakenly thought he was using ^ in that manner. So snibgo is correct, if all you want is <, then in unix, leave off the ^ or replace with \.

snibgo wrote:

Rectangle:1x100+0+0^<

In Windows, "^" is the escape character. The Windows command interpreter sees this, and passes the next character "<" as a literal to IM, instead of interpreting it as file redirection.

You don't want "^" on non-Windows platforms. You might substitute the appropriate escape character, thus:

Rectangle:1x100+0+0\<

I tried replacing the escape character on windows "^" with the one for non-windows platform "\" and now works fine also on mac osx.

Thank you very much.

Pietro

Into the documents scanned that I will pass to the OCR, I have also some dotted lines (vertical and horizontal).



Is there a way to remove them?

I'm thinking to a process like this:
apply to the original image a -motion-blur 0x4+0 and create an image_horizontal_blured.jpg
apply to the original image a -motion-blur 0x4+90 and create an image_vertical_blured.jpg

At that point, will be necessary to use some command to recognize the horizontal lines area in image_horizontal_blured.jpg, and fill of white this area in the original image creating an image as image_without_hori_lines.jpg.

So do another time the same operation recognizing the vertical lines area in image_vertical_blured.jpg, and fill of white this area in the image_without_hori_lines.jpg creating an image as image_without_vert_hori_lines.jpg.

Would be this approch correct? If yes, can someone help me to do that?

Is there a different way to remove dotted lines?

I think this argument is a very common problem in preprocessing image for OCR and could be interesting for other ones find a post to use as guide.

Thanks in advance

Pietro

Fred recently answered an almost identical question, giving a script to remove vertical and horizontal lines. A search should find it.