README

A simple OCR using the gamma classifier, coded in MATLAB, the function ocr is compiled to create a COM component, that is implemented in a proyect using Visual Basic 2008 to create the GUI.

The code has been designed for Windos OS, using the spell checker MS Office Word Activex to correct words, it needs MS Office installed.

Can use the Alfa Beta SVM as classifier too.

The component where the class ocrMain is implemented takes seven parameters, all are string datatype.

string relacion: Is the parameter used to normalized the image of the characters, results of takes the cocient of the width/hight

string patxclase: The number of patters per class that have the fundamental set or database

string nclases: Number of classes in the database, the complete set of letters that could be recognized

string umbralinc: An integer used to improve the performance, while the image is binarized

string altoinc: The incremente used to take the height-average in all the textlines contained on the image.

string normalto: The height of the normalized image crop that contains the character

string normanchi: The width of the normalized image crop that contains the character

string ruta: The path to the image to recognize

string salida: The path to the salida.txt file, with the text recognized

string base: Name of the database for the classifier, use some of the csv files in the repo

For the best performance on the recognition set the parameters with these values:
relacion="1.3"
patxclase="6"
nclases="80"
umbralinc="17"
altoinc="8"
normalto="40"
normancho="29"
ruta="path to the image input file"
salida="path to the output file"
base="courierarial40x29.csv" 

To compile the COM component use the .prj file on the Matlab Compiler.

This code has been created by Jesus Lopez.