Optical Character Recognition (OCR)

Hi, this is Diego Barragán, from Ecuador.

I appreciate your comments about OCR.

The error perhaps is due the version of MATLAB. I use version 7.1.0.246 (R14) Service Pack 3.

And other restriction of OCR is the size of letter. The size must be bigger or equal to 42 x 24 pixels.

Regards.

Nice work. Tried creating my own images of numbers and names alike and found sometimes some letters being wrongly identified, especially I, 1, O,0,Q
Other than that nice work!

Dear Sir, I'm doing handwritten english character recognition.Please guide me to progress this project. thanks lot.

hi author
i am using matlab 6.5 version. Can i work your project my matlab?

these files really works..
result not being 100%..

cz

It works great.But for images with dots or small patches the result is ambiguous.If the code can eliminate this problem,(not recognizing non characters) it will be a good algorithm for character recognition.

this program is not working for other image files. i created some image files using paint in ariel and times new roman font. the program failed to give correct result

good job!

thank you

Sid H, come on!

If you can do better, we are waiting!

dear sir,
it was a very nice job. however for certain pics it cant read the image correctly. do u have any idea on how to add the image filtering codes to make this program read the image more accurate?
i really need ur help on this matter.

I hava a question for Mr Diego Barragán.
 In the Zip file I cannt find the function clip, but On this page link http://www.matpic.com/MATLAB/MATLAB_OPTICAL_CHARACTER_RECOGNITION.html, exist!!!! Where is integrate this function?

Thank you for answare, but I have a nother question : in lines.m file exist next line:
im_texto=clip(im_texto);
What exactly make cilp in this expresion?

Hi Diego..
dis is priyanka,from India
It was a very nice job.Can u plz help me with dis project..???
Plz tell me how to recognize lower case letters..?? and also plz help me with how to obtain the bitmap images of the letters u place in letters_numbers folder so dat I can try for the regional languages...
Plz plz plz do reply me....

Problem solved. The problem was the performance of my PC.

Thanks again for the program

not very good for car plate in China.

Dear sir,
how do we detect blank spaces between the characters?
and print them in the output

Hi Diego and everyone,
Good program, but I have a few questions:
Where/how did you make these images of the letters that are in the template? I need to be able to identify lower case letters as well as upper case and have tried just using microsoft paint but although they are .bmp the appear the wrong format. Any ideas anyone?
Also - what is 'clip'? my matlab doesn't seem to support it but when i changed it to imcrop it stalls.
Thanks,
Hannah

Hi Hannah,
these images of letters are in the letters_numbers folder. there are 36 image file, if u also want to recognize lower case letters u can add lower letter image files into that folder. and need to edit and add some coding in .m file in OCR folder.
clip is not built-in function, it is a sub function use in line.m function to crop the lines, u can see it lower part of line.m function.
good luck,
thaw

sir, nice your OCR project...i also using your code for my character recognition for my project...but i code some error because of edge of image...can sir look my project and suggest a solution...i need ur help sir..tq

here is my project

http://www.filehosting.org/file/details/203684/my_project.rar

It is perfect code, really thanks

Thx!
It helps me so much!

Quite good!

plz anyone can give me the logic behind this program....on wht basis feature extraction has been done...nd hw Image proceesing tool box is used for simulatin...plz help me out asap....

i use matlab 6.5,wheh i run oCR project,then templates loading error is come.why error is come?how to error is removed?

Hello sir
It is perfect code, really thanks
Please send this project pdf

nicely done. for more robust results i found preforming a skeletal reduction of the input image text (when properly prepared), and making a template of skeletal comparisons for the correlation coefficient calculation can get around most font issues.
The next thing is properly aligning the input letter to that of the template. This can be accomplished in many ways, or you can simply preform an image rotation of all 360 degrees at take the minimum confidence coefficient of that process.