how to separate symbols from each other

[ST-Peterburg]

Hello all!

Can Anyone take me idea to solve my problem:

I must read text from image, but symbols on image  "cohere" with each other and OCR function read 2 symbols as one. Symbols has different width, so I cann't define  

one and same width for character on ROI.

I know that this problem may be solved with different filters(Laplacian,Gaussian...) and Morphology functions. 

Please, tell my how to do it.

Thanks!

[muks]

First is there a chance of getting a better uniformly illuminated image? Lemme work on the image and get back.

[ST-Peterburg]

If I have same image?:

[The_Chancenator]

Hi,

Do you have a character set file for this image? It might make it easier for others to troubleshoot. 

Thanks, 

[ST-Peterburg]

Thank You for your answer!

it is my character set file

[The_Chancenator]

Hi,

I'm thinking that we just need to "train" it a little better.  When I do the OCR example, I get a bunch of A's and ?'s, but I'm not sure if this is because of the program or the fact that I don't have Russian (they are Russian, correct?) characters on my computer.  Try going through the Vision Assistant and doing this (Programs> National Instruments>Vision Assistant).  Just open up your image and choose OCR/OCV.  Choose "edit character set" and you should be able to train your character set file.  If you directly select the ROI as the character and it doesn't recognize it, this is most likely your problem.  If it does recognize these characters, then maybe we should take a look into adjusting the image itself. 

Cheers, 

[ST-Peterburg]

Thank You!

OCR text such that is very difficult indeed.I'll try to train my character set file and use filters to separate symbols.

I wanted to know what kind of filters or morphologic function are best to do it?

[The_Chancenator]

Hi,

If training your character set a little more doesn't work, then we could try doing a Find Shapes to separate out the letters.  Also, this paper looks a bit like an alphabet; in your actual application, are the letters also going to be this close?  If not, then we probably don't have to worry about it.  If so, then we do need to go down this road. 

Please let me know how things go. 

Cheers, 

[ST-Peterburg]

Thanks for Your support!

It is really Russian alphabet on image. Text which I try to read with OCR function look like this alphabet. I have selected letter's font and size such as text(which I try to read) and I have trained my OCR function with this alphabet.

To read this text I must to separate symbols and do Horizontal element spacing bigger.And now I am trying to do it.

best regards!

[The_Chancenator]

OK, so try maybe doing a shape find and see if it can discern between the closely spaced letters.