Hi everyone, I'm new in using opencv and I need your help. My aim is to develop an Arabic OCR program for Handwritten. What will be my best way to go for? Will it be good to use one of the machine learning methods?
For Arabic ICR engine you need to perfrom the following step:
In the ICR field for train each word in the first have to normalize the word.One of the method for normalization is moments-base normalization that use the first moments to center of gravity & second nu for size of normalization.
One of the best method for Arabic handwriting use skeleton for thinning the stroke.
In the next step you have to extract the terminal and intersection point.The best method for this purpose convert skeleton to graph.
In the final step you have to train samples with hidden markov model.In this step each terminal or intersection point consider the state.
Thanks Mostafa for your reply. I tried word segmentation using skeleton and chain code but in Matlab environment. I was thinking of using the number of holes and the number of dots as features to be extracted. But, I just struggle in getting the concept of training data and testing data. Would you explain to me in simple terms what would hidden markov models need from me?
Appreciated,,
Hi you can refer the answer here might be helpful...
And see OpenCV documentation about OCR of Hand-written Data using kNN.
Also see this Blog describing about OpenCV OCR using handwrite letters.
Asked: 2013-12-22 07:00:30 -0600
Seen: 2,555 times
Last updated: Dec 22 '13
