1. A method for training of character models for a plurality of characters for optical character recognition (OCR), the method comprising:
storing at least one character model for each character in a database, each character model being trained on a first set and a second set to provide a probability of occurrence of a character in an image data wherein the first set contains a set of images of the character associated with a respective character model, and the second set contains a set of images of characters other than the character associated with the respective character model; and for each character model of the plurality of character models,
selecting a plurality of images from the first set of a character by:
generating clusters from the set of images in the first set using a clustering algorithm;
on prior presence of a character model corresponding to the character, ranking clusters of the first set based on an average probability of occurrence of images in each cluster and on absence of a character model corresponding to the character, providing a predefined rank to each cluster; and
selecting one or more images from each first set cluster weighted by rank of the cluster, selecting a plurality of images from a second set of the character by:
generating clusters from the set of images in the second set using a clustering algorithm;
on prior presence of a character model corresponding to the character, ranking clusters of the second set based on an average probability of occurrence of images in each cluster and on absence of a character model corresponding to

the character, ranking each cluster using a distance metric from the first set cluster; and
selecting one or more images from each cluster in the second set weighted by rank of the cluster.
training a new character model for the character based on the selected images from the first set and second set of the character; and
updating the database with the new character model.
2. The method as claimed in claim 1 wherein the image data is obtained by applying an OCR process to a text of a document.
3. The method as claimed in claim 1 wherein the database is updated with the new trained model on absence of a prior model corresponding to the character.
4. The method as claimed in claim 2 wherein the database is updated with the new trained model if a performance of the new character model is greater than performance of a prior model.
5. The method as claimed in claim 1 wherein the first set contains a set of samples of the character of varied font, style and size.
6. A system for optical character recognition (OCR) for a plurality of characters using a plurality of character models in a database, the system comprising:
a data store configured to store at least one character model for each character in a
database, each character model being trained on a first set and a second set to provide a
11

probability of occurrence of a character in an image data wherein the first set contains a set of images of the character associated with a respective character model, and the second set contains a set of images of characters other than the character associated with the respective character model; and
an image selector module configured to perform:
for each character model of the plurality of character models,
selecting a plurality of images from the first set of a character by:
generating clusters from the set of images in the first set using a clustering algorithm;
on prior presence of a character model corresponding to the character, ranking clusters of the first set based on an average probability of occurrence of images in each cluster and on absence of a character model corresponding to the character, providing a predefined rank to each cluster; and
selecting one or more images from each first set cluster weighted by rank of the cluster, selecting a plurality of images from a second set of the character by:
generating clusters from the set of images in the second set using a clustering algorithm;
on prior presence of a character model corresponding to the character, ranking clusters of the second set based on an average probability of occurrence of images in each cluster and on absence of a character model corresponding to the character, ranking each cluster using a distance metric from the first set cluster; and
selecting one or more images from each cluster in the second set weighted by rank of the cluster; and

a training module configured to train a new character model for the character based on the selected images from the first set and second set of the character.
7. The system as claimed in claim 5 further comprising updating the database with the new character model.